DEGEMINATION AND IN INUTTUT:

AN ACOUSTIC STUDY

by

Paul Pigott

B.A., University of Western Ontario (1995)

B.., University of King' College (1996)

A Thesis

Submitted to the School of Graduate Studies in Partial Fulfilment of the

Requirements for the Degree of

Master of Arts

Department of Linguistics

Memorial University ofNewfoundland

June 2013

St. John's Newfoundland and Labrador ABSTRACT

In this thesis, I investigate the acoustic expression of Schneider's Law (SL), a de rule observed in three dialects of (Labrador Inuttut, Quebec

Inuttitut, and Northwest Territories ), the characterization of which has so far been based largely on aural-impressionistic data transcriptions. Given the expression of this rule, which conditions alternations between that end with versus , thereby affecting rhythmic qualities of the language, I set out to perform instrumental measurements of spontaneous and elicited speech recorded in Labrador,

Canada. My observations of SL from various acoustic viewpoints confirm its characterization in the scientific literature as a virtually exceptionless rule, and one that is consistent with dissimilation processes found in other languages - in particular, length contrasts attested in Latin, Japanese and Finnish. SL is further shown to operate independently from any system of recurring metrical . Labrador Inuttut itself seems to be devoid of any type of metrical conditioning in any of the standardly-assumed phonetic correlates of stress (intensity, duration or pitch). The acoustic results do, however, show a systematic pattern of phrase-final lengthening, optionally co­ varying with FO boundary tones. The observed phenomena are consistent with descriptions of related dialects, except that the rule in Labrador Inuttut is shown to also include aspiration of phrase-final stop consonants, something not mentioned in the

Eskimo-Aleut literature, but described in other languages (.., right-edge fortition in

Blackfoot). Master of Arts Thesis

DEGEMINATION AND PROSODY IN LABRADOR INUTTUT:

AN ACOUSTIC STUDY

Presented by

Paul Pigott, B.A., B.J.

Supervisor______

Dr. Douglas Wharrarn

Supervisor______

Dr. Yvan Rose

Memorial University ofNewfoundland

201 2

11 Acknowledgements

Nakummik to thank. . Nakumminiakkat. Make sure you thank her/him. NakutlakKutit. Thank you very much.

The above is a lexical entry in UKauset katitsutauppaliningit (literally, 'a collection of words') compiled in 1976 from a weekend gathering oflnuit Elders in Nain, Labrador on 's northeast coast. It is the single most important text in the writing of this thesis and sparked my interest in the dialect of Inuktitut spoken in Labrador. A tattered copy became my virtual appendage from 1999 when my family decided to exclusively speak in Inuktitut to our newborn daughter Anika Eta Nochasak-Pigott. The simple format, 3,000 lexical entries with definitions, examples and miniature hand-drawn illustrations, was a superb introduction to the frighteningly difficult written appearance of this . The genius of the 'word collection' is the defined examples, extracted and edited by Rose Jeddore (Pamak) from the spontaneous speech of Elders who got together to talk about their language. My first thanks is to them, nakutlakKutit.

The next thanks goes to my best supporters and academic supervisors Dr. Doug Wharram and Dr. Yvan Rose. Without their knowledge and steadfast encouragement this important research could not have been realized. Thanks as well to Dr. Carrie Dyck for daring me to "crack Schneider's Law," to Dr. Paul De Decker for advice on phonetics, to Dr. Phil Branigan for taking me up the faraway tree of syntax, to Dr. Julie Brittain for her elegant insights into and to Dr. Marguerite MacKenzie for all manner of support, both academic and personal. To the wisdom of Ruby Bishop in particular, and all the staff and students in the Linguistics Department at Memorial in general, I am forever indebted. Thanks specifically to Linguistics students Karen Hunter, Kelly Burkinshaw and Victoria Allen for their transcription work. Thanks as well to Derek Norman at Memorial's Media Centre for use of professional recording equipment, Dr. Yvan Rose at Memorial's Speech Sciences and Language Acquisition Laboratory for use of the studio, and to the OKalaKatiget Society, the Inuttut broadcaster in Nain, for loans oftheir studio space, video equipment and reporter Josh Pamak.

Thanks go also the ISER Foundation, Arctic Studies Center at the National Museum of Natural History in Washington, D.C., Vale INCO, Nalcor, and Air Labrador for in-kind support and funding.

Nakummesuak, a huge thanks, to the Inuit of Labrador, the participants, translators and helpers with my field research, in alphabetical order: William Anderson

lll III, Joe Atsatata, Joan Dicker, Augusta Erving, Amos Fox, Doris Flowers, Edward Flowers, Bertha Holeiter, Boas Kaitok, Maria Harris, Julius Ikkusik, Lucas lttulak, Sara Ittulak, Veruna Ittulak, Markus Kojak, Boas Jararuse, Paul Jararuse, Christine Nochasak, David Nochasak, Levi Nochasak, Phillip Abel, Phyllis Maggo, Jacko Merkuratsuk, Minnie Merkuratsuk, Hubert Paliser, Derek Pottle, Eddie Pottle, Andrew Piercey, Martin Jararuse, Kitora Tuglavina, Mary White, Brian Williams, Fran Williams, Henry Winters, Abia Zarpa and three language consultants who asked to remain anonymous.

Finally thanks to my wife Helen Toozer. Without her love and support none of the following would be possible. Diolch, i fy nghariad.

IV Table of Contents

Abstract...... i Acknowledgements ...... iii List of Figures ...... ix Chapter 1: Introduction 1.1 Preliminaries ...... I 1.2 Thesis objectives ...... 2 1.3 The - family ...... 2 1.3 .1 Background literature on the phonology of Labrador Inuttut...... 4 1.3 .2 Thesis roadmap ...... 6 1.4 Theoretical framework ...... 7 1.4.1 Autosegmental phonology ...... 9 1.4.2 CV tier: Representing duration ...... 9 Chapter 2: Phonological Sketch of Labrador Inuttut 2.1 Accounts of the phonemic inventory and phonological processes ...... 12 2.2 Schneider's Law: A basic description ...... IS 2.3 Preliminary results ...... 20 2.3 .1 SL is exceptionless ...... 21 2.3.2 Syllable prominence is not based on loudness, or loudness and duration...... 22 2.3.3 Syllable prominence is not based on duration alone ...... 25 2.3 .4 Syllable prominence is not based on pitch ...... 26 2.3.5 Syllable prominence is not based on articulatory quality ...... 26 2.3.6 Syllable prominence is not based on pitch and duration ...... 27 2.4 Summary...... 27

v Chapter 3: Background Literature on Schneider's Law, "stress" and pitch 3.1 Introduction ...... 28 3.2 SL descriptions ...... 28 3.2.1 SL in Quebec ...... 29 3.2.2 SL in the ...... 35 3.2.3 SL in Labrador Inuttut...... 37 3.2.4 Deleting versus adjoining affixes in Labrador Inuttut...... 39 3.3 SL analyses ...... 41 3.3 .1 SL is not "geminate tension" in Quebec Inuttitut ...... 41 3.3 .2 SL is not a metrical phenomenon ...... 44 3.4 Accounts of"stress" ...... 53 3 .4.1 Early accounts of "stress" in ...... 54 3.4.2 A recent account of "stress" in West Greenlandic ...... 58 3.5 Accounts of pitch in the Inuit languages ...... 59 3.5.1 Accounts of pitch in West Greenlandic ...... 59 3.5.2 A recent account of tonal and durational patterns in West Greenlandic ...... 61 3.5.3 Pitch patterns observed in Quebec Inuttitut...... 63 3.6 Summary...... 69 Chapter 4: Methodology 4.1 Preliminaries ...... 70 4.2 Participants...... 70 4.3 Audio environment and equipment...... 71 4.4 Goals and methodology ...... 71 4.4 .1 Design and implementation of the linguistic interview ...... 72 4.4.2 Ethnographic interviews: Purpose and methodology ...... 74 4.5 Software and settings ...... 76

Vl 4.5.1 Acoustic Analysis ...... 77 4.5.2 The phonemic pair / tutuk/ ' messy hair' vs. / tuttuk/ •caribou' ...... 77 4.5.3 The / xattax/ ' often, intermittently' ...... 78 4.5.4 The sequence /xixxi/...... 79

4.5.5 The sequence /na:na/...... 80 Chapter 5: Results and analysis 5.1 SL: Acoustic manifestation ...... 81 5 .1.1 Fieldwork data consistent with previous descriptions of SL...... 81 5.1.2 SL holds despite variations of the 3s negative morpheme ...... 86 5.1.3 SL operates independently from vowellength ...... 88 5.1.4 Summary of durational results for /tutuk/ versus / tuttuk/ ...... 90 5.2 SL: No metrical motivation ...... 92 5.2.1 SL: Unrelated to syllable intensity ...... 93 5.2.2 SL: Unrelated to ...... 96 5.3 No evidence of metrical stress in the data ...... 107 5.3 .1 Intensity prominence does not depend on duration ...... 107 5.3.2 Intensity prominence does not depend on prosodic factors ...... 111 5.4 Evidence for syllable timing ...... 115 5.4.1 Syllables in example words are similar in length ...... 116 5.4.2 Syllables in phrases are similar in length ...... 119 5.5 Summary...... 126 Chapter 6: Conclusion ...... 127 Bibliography and References ...... 131

Vll APPENDICES A: Linguistic interview ...... 142 B: Ethical approval ...... 151 C: The phonemic pair / tutuk/ 'messy hair' versus / tuttuk/ 'caribou' ...... 152 D: The morpheme / x_attax/ 'often, intermittently' ...... 155 E: The sequence / x_ixxi/ ...... 159 : The sequence / na:na/ ...... 166

Vlll List of Figures

Figure 1: Slightly falling ...... 98 Figure 2: HLH boundary melody ...... 100 Figure 3: HLH boundary melody ...... 101 Figure 4: HLH boundary melody ...... 103 Figure 5: HL boundary melody ...... 105 Figure 6: Intensity prominence is discretionary ...... 110 Figure 7: Evidence of a declarative phrase ...... 112 Figure 8: HL boundary tone ...... 112 Figure 9: HL boundary tone ...... 113

lX 1 Introduction

1.1 Preliminaries

This thesis is about a degemination phenomenon described in Eskimo-Aleut literature as

Schneider's Law (henceforth, SL). Through instrumental measurement of Labrador

Inuttut spontaneous speech, I will show that SL applies exceptionlessly throughout hundreds of spoken utterances. Following an insight from Dresher & Johns (1995Y, who state that SL cannot be related to metrical stress, I show that SL does not co-vary with any measurable phonetic correlate of stress. Indeed, I show that SL is unrelated to any rhythmic or intonational characteristic of the system. For example, I show that lengthened phrase-final syllables co-vary with /declarative boundary tones, without regard to the application or non-application of SL. The results are thus consistent with acoustic studies of related dialects (e.g., in West Greenlandic ), which also find no evidence of a metrical system of alternating stress in the language and (e.g., in Quebec

Inuttitut) exceptionless degemination of underlying geminate consonants in syllable- adjacent positions moving from left-to-right. Finally, SL is typologically consistent with dissimilation processes in other languages. Bye (2011 :3) describes fifteen types of dissimilation; relevant here are the languages that use the suprasegmental property of length as the medium for syllable-adjacent contrasts including Finnish (Keyser &

Kiparsky 1984), Gidabal (Geytenbeek & Geytenbeek 1971 ), Japanese (Iwai 1989, Wade

1996, Ito & Mester 1998), Latin (Leumann 1977, Sihler 1995 and Ito & Mester 1998),

' The law of double consonants in Inuktitut' was first presented at the 35' annual meeting of the Canadian Linguistics Association, in Victoria, B.C. May 27-29, 1990.

1 Oromo (Gragg 1976, Lloret 1988, Alderete 1997) and Slovak (Kenstowicz & Kisseberth

1977, 1979 and Rubach 1993) .

1.2 Thesis objectives

In this study I attempt to document the SL phenomenon in current speech samples from

Labrador Inuttut language consultants. The goal is to verify proposals found in the previous scientific literature on SL, especially Dresher & Johns ( 1995), who argue that the rule is virtually exceptionless, refers exclusively to underlying geminates in adjacent syllables and operates independently from duration, intensity, and pitch prominence. To confirm the latter, a secondary aim of this study is to show the nature of intonation in current Labrador Inuttut speech with the expectation that the data studied here will be consistent with the accounts of intonation in West Greenlandic (Mase 1973, Rischel 1974 and Nagano-Madsen 1990, 1993, 1994) and Quebec Inuttitut (Massenet 1980). The final goal of this thesis is to say something about the nature of rhythm in Labrador Inuttut, an issue that remains unresolved in the scientific literature on the Inuit languages.

1.3 The Eskimo-Aleut

The Eskimo-Aleut Language Family is a continuum of related grammars and vocabularies which includes six languages spoken on the Aleutian Islands, a dialect grouping spoken on the Chukotka Peninsula and in Southern , and a group of mutually intelligible dialects spoken across the North American Arctic from the Bering

Strait in Alaska to the East coast of and Southern Labrador (Dorais 1990b).

2 The intermediate group comprises the various dialects ofYupik, which Hayes (1995:239) describes as stress-timed with persistent left-to-right iambs, and with the context­ dependent assignment of either one or two moras to CVC. The latter group is the

'Eskimo' branch of the family and includes Inupiaq in Alaska and the Northwest

Territories, and Inuktitut in Canada as well as several dialects in Greenland.

Dorais (1990b:2) calls the Eskimo branch the ' Inuit language' and that is how this group of related dialects will be referred to here. Dorais claims that speakers from as far away as the Bering Strait and Labrador can, with some difficulty, understand each other. Across the Inuit languages, Creider ( 1981) observes a continuum of Regressive , showing that from the perspective of phonology, each dialect is slightly different in the way it deals with heterogeneous consonant clusters. Following the schematic generalization in Dorais (1990:41), the most conservative form is Western Alaskan

Inupiaq in which there are eleven attested underlying environments for consonant clusters at the surface level: [mr, nr, XC, UVULARC, kC, VELARC, pC, BILABIALC, tC, ALVEOLARC, jFRICATIVE] . In the dialects oflnupiaq spoken in the North Slope region of Alaska and

Northwest Territories, the number of environments drops to eight with the [XC, kC, pC, tC] distinctions assimilating into surface geminates. Inuktun has seven underlying environments for formation and the Western Canadian Arctic dialects of

Inuktitut have five. The Baffin Island and Arctic Greenland dialects of Inuktitut have four, the same number as the remaining two dialects in Greenland. The dialect of Inuktitut in

Quebec has only two environments, [riJ, UVVLARC], illustrated for example by the fact that speakers call their language, phonetically, [inuttitut]. Dresher & Johns (1995 :83) show

3 that the process of RA is complete in Labrador lnuttut, an Eastern Canadian dialect of

Inuktitut where consonant clusters assimilate for manner, primary place and secondary place. The primary data collected for this thesis come from 31 Labrador Inuttut speaking language consultants. It is of primary interest as one of only three dialects in Eskimo-

Aleut shown by Smith (1975:105) to have exceptionless degemination so that, moving left-to-right, the second in a pair of syllable-adjacent geminates is always reduced:

CV(V)CCV(V)CCV---+ CV(V)CCV(V)CV. The phenomenon was first documented by

Father Lucien Schneider, a Catholic missionary who lived with Inuit in Quebec from the late 1930's until his return to France in 1974. Schneider (1966) describes the pattern for

Quebec Inuttitut while Dorais & Lowe (1982) show a more restricted form of SL in the

Inuktun dialect ofNorthwest Territories Siglitun.2

1.3.1 Background literature on the phonology of Labrador Inuttut

The compiled works on Labrador Inuttut phonology make up a slender volume in the

Inuit languages literature, a body of linguistic inquiry that begins in Greenland with a phrase book by (1721) and the dictionary of his son Poul Egede (1760:6-7).

The latter shows that length is phonemic with a series of minimal pairs. Rischel

(1974:26) compares phonemic pairs from current West Greenlandic with their counterparts in Egede (1760), showing how vowel length is phonemic:

2 In Northwest Territories Siglitun the schema CV(V)C; C;V(V)C;C;V(V)-> CV(V)C; C;V(V)C;V(V) properly describes the behaviour of SL. However unlike Quebec Inuttut, where SL also targets underlying consonant clusters, in Northwest Territories Siglitun the rule applies exclusively to underlying geminates. As a result the following schema is attested in that dialect: CV(V)C; C;V(V)C;Ci V(V)-> CV(V)C; C;V(V)C;CiV(V).

4 (1) West Greenlandic (Egede 1760) West Greenlandic (Richel 197 4) a. aularpok3 /aularpok [a:la:rpuq] I [ a:larpuq] ' moves' 'leaves'

b. manna/ manna [ma:nna]/[manna] ' now' 'that one'

The opaque orthography used by Egede in (1) makes it difficult to interpret the accent marks used to show vowel length, but Rischel 's comparison shows that they must relate to phonemic vowel contrasts. Rischel (1974) makes another comparison from Egede

(1760) to current West Greenlandic, showing that phonemic contrast can also be based on consonant duration:

(2) West Greenlandic (Egede 1760) West Greenlandic (Richel 197 4) a. frsilerpalirsilerpa [issilirpa:]/[isilirpa:] ' begins to freeze' ' begins to look at him/her'

b. aggiuta/aggiuta [ayyiiuta:]/[ayiiuta:] ' day of arrival' ' his/her file'

Rischel (1974:91) writes that the "terminology [used by Egede] reflects a transfer from

Latin grammar and metrics rather than a real analysis of the Eskimo pattern, but it happened to be the crucial distinction in Eskimo as well." These phonemic contrasts are one hallmark of the Inuit languages; acoustic results shown in §5.1.1 document this phenomenon in current Labrador Inuttut speech. The literature also describes most of the

Eskimo-Aleut languages in Alaska and Siberia as stress-timed and iambic. Only a handful

3 Egede's use of the mid tone marker in this example and the hat marker in (I b) seems to indicate in both examples that the vowel is long.

5 of phonological studies have been done in the Central and Eastern Arctic, with published acoustic results only for West Greenlandic and Quebec Inuttitut. Kleinschmidt (1851) contains an aural-impressionistic account of West Greenlandic phonology. Instrumental measurement for this dialect comes from Mase & Rischel ( 1971 ), Mase ( 1973 ), Rischel

(1974), Nagano-Madsen (1990, 1993, 1994), and Jacobsen (2000). The latter concludes that stress is not a relevant category, based on her measurement of words produced in carrier sentences by two language consultants. Massenet (1980) derives a similar conclusion from his acoustic analysis of Quebec Inuttitut. As there are no acoustic studies of Labrador Inuttut to follow, these accounts of West Greenlandic and Quebec Inuttitut inform the analysis of the data considered here. This thesis will not rely on Erdmann

( 1864) or Bourquin (1891 ). Both adopt Kleinschmidt's orthography and phonological account of West Greenlandic, impacting the written form of Labrador Inuttut and confusing the phonological picture of the dialect until Dresher & John's (1995) account, a work which forms the basis of the phonetic and phonological description of Labrador

Inuttut in Chapter Two.

1.3.2 Thesis roadmap

In the remainder of this chapter I present the theoretical framework for this thesis.

Chapter Two begins with a phonetic and phonological sketch of Labrador Inuttut, followed by a brief description and schematic representation of SL as observed in the

Inuit dialects. I conclude Chapter Two with a discussion of my preliminary results, which show no evidence of syllabic trochees, moraic trochees or iambs as described in Metrical

6 Stress Theory (MST hereafter) by Hayes (1995). In Chapter Three, I discuss in detail the background literature on SL in the Inuit languages, syllable prominence in West

Greenlandic and intonation in West Greenlandic and Quebec Inuttitut. Chapter Four contains a discussion of the methodology used to design the linguistic interviews, implement the fieldwork, and analyze the results. The end of that chapter includes a detailed description of four types of data and the manner in which they were analyzed to produce the results introduced in Chapter Five. That chapter summarizes results which show that SL is virtually exceptionless, unrelated to metrical conditioning, and independent of intonation. I conclude in Chapter Six with a brief summary of the most central arguments of this thesis as well as some of their contributions for the field and related areas.

1.4 Theoretical framework

This thesis follows the model of multilinear generative phonology, assuming segments are made up of distinctive features. These largely coincide with the feature bundles proposed in Chomsky & Halle's (1968) Sound Pattern ofEnglish, but that linear approach is rejected here for a model where the various units comprised within phonological systems can be divided into independent components with potentially multiple associations between them. Goldsmith's (1976) autosegmental approach recognizes that features can spread across segmental or prosodic boundaries, grouping this sharing of segmental material into a geometry of interdependent features, the relations between them forming the locus of phonological patterning. This thesis adopts

7 the skeletal tier to represent segmental duration. For syllabification, consider Kahn's

(1976) Maximum Onset Principle (MOP), stated as follows:

(3) Maximum Onset Principle (MOP) (Kahn, 1976) First make the onset as long as it legitimately can be; then form a legitimate coda.

This thesis follows Kahn's view that the context of phonological rules can often be captured by referencing duration to syllable-based generalizations. The rhythmic model accepted here is Selkirk's (1980) hierarchically organized prosodic domains, including the prosodic word, foot, syllable and . For the latter, this thesis follows Prince (1984) in representing intervocalic geminates as doubly-linking consonantal material to both a coda and the onset position following it across the syllable boundary. While none of the data produced by the current research contradicts the generalizations stated within MST as proposed by Hayes (1995), this thesis shows that Labrador Inuttut is not a "stress- timed" language, and therefore lies outside the MST framework. This thesis follows, instead, the syllable timing alternative proposed by Kager (1993, 1995) and "syllable- timed" theory as proposed in Abercrombie (1967), Ladefoged (1975), Roach (1982),

Ramus, Nespor & Mehler (1999) and Mehler, Nespor, & Shukla (2011). Each ofthe above will now be considered in more detail.

8 1.4.1 Autosegmental phonology

A multilinear approach to features allows a better description of assimilatory processes.

As mentioned in § 1.3, in the Eastern Canadian dialects of the Inuit languages, consonant clusters undergo Regressive Assimilation, so that [inuktitut] 'like the Inuit' becomes

[inuttitut] in Quebec Inuttitut: dorsal-coronal clusters become coronal-coronal / kt/ ---+ / tt/ .

Under SPE, general rules would result in an inelegant description that depends on a self­ contained list of place features. In an autosegmentalized representation, segmental features are organized under nodes for manner (, nasal and continuant) and place, with assimilation spreading featural material from one node (the trigger) to the next (the target). Dresher & Johns (1995) use feature geometry (see §3.3.2) to account for the natural classes of sounds that participate in these assimilation processes.

1.4.2 CV tier: representing duration

From the articulatory characteristics of segments, we tum next to the representation of their length. The skeletal tier organizes segments into an abstract sequence of time units or slots which can then have the added specification of[± syllabic], with vowels (V) being [+syllabic] and consonants [-syllabic]. I adopt the CV-Tier model argued for by

Clements & Keyser (1983), so that the examples [inuktitut, inuttitut] can both be represented as VCVCCVCVC or, if we treat geminates as taking up a single slot,

[inuttitut] = VCVCVCVC. The of geminate representation is a key issue in the study of Labrador Inuttut where all geminates are underlyingly clusters of consonants and thus both the target and trigger of SL. CV-Tier theory can be of assistance in explaining

9 geminate structure because it allows for slots with no segmental material or with segments not associated with a single syllable node. In this study, I will consider syllable- based generalizations that might help us to represent SL in Labrador Inuttut. The dichotomy between /kt/ and /tt/ as CC is thus represented in ( 4), where the geminate in

( 4b) straddles two syllables and is not limited to one syllable node:

(4) CV-Tier representation in Inuktitut. a. cr cr !\;1 e vee v I I I I I u

b. cr cr !l\ ;1 eve ev II VI n u t

Syllabification in Inuttut follows the Maximum Onset Principle, stated in (3), above. The fact that !kt/ and / tt/ are not legitimate onsets supports syllabification as

V.CVC.CV.CVC, with two timing slots for geminates. Further support comes from the feature geometry posited by Dresher & Johns (1995:88), where pharyngeal geminates in

10 Quebec Inuttitut are shown to have two nodes. The implications for Labrador Inuttut and the data considered here are the of the next chapter.

11 2 Phonological sketch of Labrador Inuttut

2.1 Accounts of the phonemic inventory and phonological processes

The first mention of Labrador Inuttut phonology in the literature comes from published letters between Kleinschmidt & Bourquin (1881). Their discussion centres on the orthographic representation of Labrador Inuttut versus the dialects in Greenland.

Kleinschmidt's system was adopted in Labrador by Moravian Missionaries, resulting in a written form that did not match the phonetic reality of the spoken language. This led to a religious form of the dialect heard today only in church, "Moravian lnuttut," which will not be considered in this thesis. The data under investigation come from linguistic interviews, including directed oral tasks, reading, descriptions of images and spontaneous conversations with language consultants in informal settings. The first generative phonological description of Labrador Inuttut comes from Smith's (197 5:1 01)

"autonomous phonemic inventory," adapted in Smith ( 1977a), Dorais ( 1990b, 2003) and

Dresher & Johns (1995), as follows:

12 (1) Labrador lnuttut consonant inventory (adapted from Dresher & Johns 1995 :82) Labial Coronal Palatal Velar Uvular Glottal voiceless stops p t k *

voiceless s X X h voiced fricatives v/{3 y lateral approximants i, 1

nasals n I) N glides j * / q/ varies with the velar / k/

This inventory is consistent with the consonants observed in the data considered here, though [{3], [N] and [h] are virtually absent. Also, in the context of phrase/utterance-fmal lengthening and pitch effects detailed in §5.2, the stops [p], [t], [k] and [q] are in complementary distribution with [ph], [th], [kh] and [qh]. These aspirated variants arise in the rightmost segment position of a phrase or utterance, while the unaspirated stops [p],

[t], [k] and [q] arise elsewhere.

All of the stops and most of the fricatives from the inventory in ( 1) are shown by

Smith (1975) as having a phonemically long, meaning-changing, variant:

(2) Labrador Inuttut surface geminate inventory (adapted from Smith 1975:102) pp tt, ti kk qq kx QX

ff,pv XX, XX H,

mm nn 1)1), NNt' d3 * / IJIJ/ varies with the uvular / NN/

13 Smith's (1975) representation ofthe mixed clusters [kx], [qx], [ts], [ti] and [d3] as is not entirely correct. Dresher & Johns (1995) show that each of these clusters is in fact the phonetic realization of geminates at the phonemic level. Part of Smith's treatment of clusters is due to the misleading orthography. 1 For example, the consonant cluster Smith writes as -kq- is, phonetically, [kx] or [qx]. These consonant clusters arise at the surface level in the data considered here and the crucial insight from Dresher &

Johns (1995) is that [kx] and [qx] derive through affrication from the underlying phonemes /xx/ and l XXI, respectively, and that [ts], [ti], and [d3] derive from the underlying phonemes Iss/, !HI and / jj/, respectively. Dresher & Johns's (1995) further show that each of these underlying phonemes is the result of Regressive Assimilation:

(3) Regressive Assimilation in Labrador Inuttut (Dresher & Johns 1995:82) X X X X

According to Dresher & Johns's (1995) analysis, if a voiceless stop [p], [t], [k] or [q] arises in the C2 position of (3), it assimilates C, resulting in the phonemes [pp], [tt], [kk] or [qq]. A voiceless [s], [X] or [i] in the C2 position also assimilates the place of articulation ofC, but the process results in affricates [ts] , [qx] or [ti] at the surface level under the rule (1995:82): "voiceless spirant geminates are affricated." The nasals are straightforward, with Regressive Assimilation resulting in the surface forms [mm], [nn],

Prior to the adoption of a standardized orthography by Labrador Inuit in 1980.

14 [IJIJ] and [NN]. The voiced fricatives [v ] and [y] assimilate C 1and then devoice, realized as [ff] and [x.x./xx] under the rule (1995 :83): "voiced obstruent geminates devoice." The

lateral approximant [1] is optionally voiceless. As a cluster it assimilates C 1 to form the underlying geminate /11/, which then affricates to [ti], following the above affrication rule.

Finally, the glide [j] assimilates C 1 to form the underlying geminate !jj!, which then affricates to [d3], following the above affrication rule. Dresher & Johns (1995) thus show that Regressive Assimilation applies to all underlying clusters in Labrador Inuttut. They also show that Regressive Assimilation must be followed by Affrication and Devoicing in the following rule ordering:

(4) Labrador Inuttut phonological processes (adapted from Dresher & Johnsl995:83) Underlying /Cp/ / Cv/ / Cs/ / Cy/ ! CX.I ! Cj! Regressive Assimilation pp vv ss yy X. X. jj (Palatal) Affrication ts qx. d3 Devoicing ff X. X. Surface [pp] [ff] [ts] [X. X.] [qx.] [d3]

Dresher & Johns (1995) also refine Smith's (1977b) generalization, based on the following data, that SL simplifies geminates and heterogeneous consonant clusters:

(5) SL and 'mixed clusters'(Smith 1977b, from Dresher & Johns 1995:83) a. pisu(k) + kqaa + vuk - pisukqaavuk 's/he walks' b. ikqa + kqaa +vuk - ikqaqaavuk 's/he remembers' c. inu(k) + atsuk - inuatsuk 'loveable inuk' d. inni(k) + atsuk - inniasuk 'loveable son'

15 As already mentioned, the ' mixed cluster' that Smith writes as -kq- is phonetically [kx] or

[qx], derived from the underlying lxx/. Smith's -ts- is derived from the underlying phoneme Iss/. Given those assumptions, Dresher & Johns (1995) show that in

Labrador Inuttut SL applies only to phonological geminates and also must follow

Regressive Assimilation. The following derivations yield the proper surface forms:

(6) SL and ' mixed clusters': sample derivations (from Dresher & Johns 1995:84) Underlying / ixxa + xxa: + vuk/ / inni(k) + assuk/ Truncation2 inniassuk SL ixxaxa:vuk inniassuk Affrication iqxaxa:vuk Surface [iqxaxa:vuk] [inniassuk] Smith ikqaqaavuk inniasuk 's/he remembers first' ' loveable inuk'

The data collected for this thesis are consistent with Dresher & Johns's (1995) analysis of rule ordering in (6). Consider the following two examples, extracted from spontaneous speech:

2 According to Smith ( 1977b:8), Labrador Inuttut has two classes of . For 'deleting suffixes', the final consonant of a base-stem is elided as in: / inuk/ 'person'+ / IJa/ '3poss' - [inuiJa]. For 'adjoining suffixes' the final consonant of a base-stem is preserved as in: / inuk/ 'person'+ / mut/ 'from the ... ' --> / inukmut/ Regressive Assimilation- [inummut] (see §3.2.4).

16 (7) SL in current Labrador Inuttut data Underlying / iylu(k) + kkut/ / axiyyi(k) + pvi + ssia + pvak/ Truncation iylukkut axiyyi pvissiapvak Regressive Assimilation illukkut axiyyivvissiavvak SL illukut axiyyivissiavak Affrication axiyyivitsiavak Devoicing axixxivitsiavak Surface [illukut] [axixxivitsiavak] 'through the house' ' big, pretty willow ptarmigan'

Vowels do not play a significant role in this thesis, except in §5.1.3 where I show that vowel length does not co-vary with intensity prominence, pitch prominence or SL.

Consider the vowel inventory in Smith ( 1977a:2):

(8) Labrador Inuttut vowel inventory Front (unrounded) Back (rounded) High [i] [u]

Low [a, a]

Smith's inventory is basically consistent with the short syllable peaks observed in the data considered in this thesis.3 While I will not investigate the phonetic realizations of the [a] and [a], I will show that there is no evidence of widespread or systematic vowel reduction. Smith (1977a:3) describes the long vowels as follows:

3 The segment (a] is virtually absent from the data considered here.

17 (9) Labrador Inuttut long vowel inventory Front (unrounded) Back (rounded) High [iu]

~~ [ui] ~[~ [~ ~[au] ~~~ [a:], [a:] Low

The data considered in this thesis are also consistent with this inventory. As we will see in

§5.2.2, a further attestation is that of overlong vowels [a::], [i:u] etc., which occur in environments where a vowel that is already long is lengthened at the right edge of a phrase/utterance.

2.2 Schneider's Law: A basic description

SL was first categorized in the literature as a weight rule that deletes the rightmost coda consonant in adjacent CVC syllables. SL effects are reported for only three Inuit dialects:

Labrador Inuttut, Quebec Inuttitut and Northwest Territories Siglitun (Schneider 1966,

Collis 1970, Rischel 1974, Smith 1975, Dorais 1976, Dorais & Lowe 1982, Fortescue

1983, Lowe 1984, Massenet 1980, 1986, Dresher & Johns 1995, 1996 and Jacobsen

2000). A straightforward demonstration of the rule comes from the following data:

18 (1 0) SL in Quebec Inuttitut (adapted from Dorais 1990b: 124) a. CCVCjCjV--+ CCVCjV illu + kkut house.vialis *illukkut illukut 'through the house' b. CVCjCjV--+ CVCjCjV nuna+kkut land.vialis *nunakut nunakkut 'through the land'

The early literature on SL describes the environment for this rule in (1 Oa) as a clash of two adjacent closed syllables. In (lOb) then SL applies vacuously. The implication ofthis position is that eve syllables are heavy. However, Dresher & Johns ( 1995) show that the presence of another kind of heavy syliable, ev:, has no impact on the operation of SL.

For example, [illu:kut] ' through two houses' and [nuna:kkut] 'through two lands' are also grammatical. First of all,"(:)" must be added to the schematic representation of SL in

(lOa) CCV(:)ejejV--+ eiCV(:)ejV. And since syllable weight cannot be a factor in the rule, SL must be redefined in Labrador Inuttut, where Regressive Assimilation applies to all heterogeneous consonant clusters, as a ban on syllable-adjacent geminates. 4

Based on his analysis of Labrador Inuttut, Smith (1977b) argues that SL applies to the first segment in ejej sequences. His evidence is based on the ' mixed cluster' he writes as -kq-, which arises as -q- when it is the target of SL. A problem with this description arises from Dresher & John's (1995) rule ordering, which posits -kq- as the phonological geminate l XXI at the point in the derivation where SL applies. The geminate l XXI

4 And syllable-adjacent underlying geminates in Quebec [nuttitut.

19 therefore says nothing about which segment in a CjCj sequence gets deleted. Smith's insight is probably correct however given the facts in Quebec Inuttitut. In that dialect, assimilation extends only as far as primary place of articulation (Dresher & Johns

1995:84) so that heterogeneous consonant clusters involving a uvular such as [qp] are possible at the point in the derivation where SL applies. SL targets these clusters, just as it targets geminates in (1 0), and in those cases the first member of the cluster is deleted, e.g.

[qp]-+[p]. Our schematic representation in (lOa) must therefore be revised again to include this possibility in Quebec Inuttitut: CiCV(:)CjCkV - CCV(:)CkV. In Northwest

Territories Siglitun, the rule operates somewhere in between the two Eastern dialects. It is like Labrador Inuttut, in that geminate consonants are targeted by SL but unlike Quebec

Inuttitut because heterogeneous consonant clusters cannot be the target of simplification

(Lowe 1984): CCjV(:)CkCkV- C CjV(:)CkV but CCjV(:)CkCV- C CjV(:)CkCV.

2.3 Preliminary results

The data considered in this thesis are consistent with the phonemic inventories and phonological rules outlined in the previous sections. In addition to spontaneous speech, language consultants participated in an oral task, designed to produce two types of example words: one with syllable-adjacent underlying geminates in all word-medial consonant positions and another with syllable-adjacent underlying geminates in all word­ medial consonant positions except the base-stem (for a full description of the methodology, see Chapter Four). The full phonetic results will be presented in Chapter

Five, but first a brief discussion of the preliminary results.

20 2.3.1 SL is exceptionless

First consider the basic description of SL above and the iterative left-to-right pattern of degemination in the following two examples:

( 11) SL in Labrador Inuttut data a. Underlying /tuttu(k) + U: + IJIJUa + x_x_au + IJIJik + tuk/ Truncation tuttu:IJIJUax_x_aUIJIJiktuk Regressive Assimilation tuttu:IJIJUax_x_aUIJIJittuk SL tuttu:IJuaxxauiJittuk Affrication tuttu:IJuaqx_auiJittuk Surface [tuttu:IJuaqx_auiJittuk]

b. Underlying / tutu(k) + U: + IJIJUa + x_x_au + IJIJik + tuk/ Truncation tutu:IJIJUax_x_aUIJIJiktuk Regressive Assimilation tutu:IJIJUax_x_aUIJIJittuk SL tutu:IJIJUax_aUIJIJituk Affrication Surface [tutu:IJIJuax aUIJIJi tuk]

SL degeminates underlying CjCj without exception in (11). This is representative ofthe results in Chapter Five for 32 examples like the ones in (11). These examples show that

SL is fully productive in the language and that it behaves just as it is described in the

Eskimo-Aleut literature. Typologically, SL is similar to a class of dissimilation rules (Bye

2011: 1285) that refer to length. Consider the Lex Mamilla rule in Latin, shown in the following data from Ito & Mester (1998:22, 70):

21 (12) Lex Mamilla rule in Latin

a. mamma ----+ mamilla *mammilla ' breast' (diminutive) offa ofella *offella ' morsel' (diminutive) saccus sacellus *saccellus 'sack' (diminutive) ob- -mitto *ommitto 'aside' 'lay aside' b. canna canalis *cannalis 'reed' 'channel' farr- farina *farrfna 'spelt' 'meal, flour' currus curUlis *curriilis 'chariot' ' relate to c.' pollen polenta *pollenta

In terms of banning syllable adjacent geminates, the rule as shown in (12a) looks just like

SL, except the output is regressive while SL applies progressively. Lex Mamilla is unlike

SL in that, as well as banning syllable adjacent geminates, a geminate in Latin also cannot arise before a heavy syllable (CV:, CVC) as shown by (12b). As will be shown in §5.1.3,

SL is a rule that targets only underlying geminates: vowel length or the metrical weight of an adjacent syllable have no impact on its application.

2.3.2 Syllable prominence is not based on loudness, or loudness and duration

Initial analysis of lexical word examples from spontaneous speech shows that intensity peaks do not pattern systematically. The observed range of intra-word intensity variation is never more than 10 decibels (dB), this value being an extreme one, as variation in loudness between syllables is typically very small. A stringent approach to this data

22 compilation results in the somewhat arbitrary cut-off of 1 dB as the minimum difference between syllables that contrast for intensity. On that basis, consider the following:

(13) a. axlxxik 'ptarmigan' AB b. axlxxfk AE c. axixxfk AE d. axixxlk BK e. axlyyfk HW f. axixxlk PJ g. axiyyl:k HP

Under this description of syllable prominence, peak intensity falls on either the initial or final syllable. Secondary prominence can fall on any syllable. Also, peak intensity and durational prominence do no co-vary. In a slightly longer word example peak intensity can fall on the penultimate syllable, and every other syllable except the initial:

(14) a. axixxllfk 'spruce ptarmigan' BK b. axixxflik BK c. axixxflik JI d. axlxxilik JM e. axixxllik SI f. axixxillk SI

The prominent syllable in terms of intensity is most often the penultimate, a generalization that gains support from another lexical word with the same number of syllables:

23 (15) a. axixxivik 'willow ptarmigan' BH b. axixxivik BK c. axixxivik EF d. axiyyivik HW e. axiyyivi:k 'is it a willow ptarmigan?' HP f. axiyyivik HP g. axixxivik JI h. axlxxivik JM 1. axixxivik LI J. axixxivi:k LI k. axixxivik PJ 1. axixxivi:k SI

As we see in (14-15), the penultimate syllable is most prominent in 12 of the 18 examples. The question is how to explain initial prominence in ( 15a, e), antepenultimate prominence in (14e) and final prominence in (14a) and (15b, d). Prosodic factors such as syllable shape do not resolve these apparent exceptions. The data in ( 13-15) come from ethnographic interviews (see full description in §4.2.2) during which language consultants were shown unlabelled photographs of Labrador flora and fauna species. In

(15), for example, the photographic plate showed a willow ptarmigan in winter and summer plumage. The spontaneous response was most often a single word answer, as in

(15a-c, g, h, k). These examples words have initial, penultimate and final prominence, clear evidence that peak intensity does not pattern systematically at the prosodic level of the word. The remaining example words are embedded in phrases and utterances, but peak intensity does not pattern consistently in these either. Observe penultimate prominence in (15f) where the example word arises phrase-initially and in ( 15j, l) where the example words are phrase-final. The initial syllable is the most prominent in (15e)

24 where the example word arises at the end of an interrogative phrase and utterance, further evidence that peak intensity does not pattern systematically at the level of the phrase/utterance.

Another initial finding is that the long vowels in ( 15) do not systematically attract peak intensity. Most accounts of Labrador Inuttut, including that of Dresher & Johns

(1995 :89), argue that long vowels attract the most "stress." In ( 15e ), the opposite can be true, with the short vowels in the initial and penultimate syllables attracting intensity peaks of72 dB and 71 dB respectively. By comparison, the long vowel in the final syllable has a peak intensity of only 68 dB. These contrasts are small enough to be considered insignificant. What truly matters is that the location of peak intensity is variable. These initial results in ( 13-15) lead to the conclusion that intensity is not a relevant correlate of syllable prominence in Labrador Inuttut (see §5.3).

2.3.3 Syllable prominence is not based on duration alone

Each vowel in Labrador Inuttut's phonemic inventory [a, i, u] has a meaning changing phonemic long form. From spontaneous speech in the data, observe the following minimal pairs:

(16) a. anak ' faeces' a:nak 'paternal grandmother' b. innik 'son' i:nnik 'starfish' c. inuk ' person' inu:k 'two people'

25 Within the word domains in (16), phonemic short vowels always have less duration than phonemic long vowels. There is no durational overlap between long and short, consistent with Jacobsen's (2000:54) acoustic study based on a reading task in West Greenlandic. As well, there is no rhythmic constraint on long vowels, which can arise in all syllable types and positions within the word. It follows from this that phonemic vowel duration alone is not relevant to any notion of stress in Labrador Inuttut.

2.3.4 Syllable prominence is not based on pitch

The Inuit dialects and all languages in the Eskimo-Aleut Family are non-tonal. This is similar to languages like English, in which tonal contrasts are not phonemically relevant, and like Japanese, where pitch is relevant at the level of the phrase and where boundary melodies are timed with the rightmost mora of an intonational phrase (Nagano-Madsen

1994). I discuss pitch effects in Chapter Five, §5.4, where I show that SL is unaffected by a pattern of pitch effects at phrase-final/utterance-final boundaries. Neither can these effects be associated with metrical stress. In non-boundary environments, pitch is slightly falling from left-to-right. Pitch alone is therefore not a relevant correlate of stress.

2.3.5 Syllable prominence is not based on articulatory quality

None of the Eskimo-Aleut languages display a systematic pattern of vowel reduction. The

Aleut and have a schwa segment in their phonemic vowel inventories.

Dialects like Siberian Yupik, for example, have [a], where the segment is unlike its [a i u] vowel counterparts in that it does not have a long form (Reuse 1994: 18). Schwa is

26 merged with [i] in all but one of the Inuit dialects. 5 I observe no schwa segments and no vowel reduction in the data considered in more detail in the next chapters.

2.3.6 Syllable prominence is not based on pitch and duration

The only pattern of pitch effects in my data involves phrase-final/utterance-final boundaries, a phenomenon that can include syllable rhyme lengthening, as shown in

Chapter Five in §5.4. Because this covariance of pitch and duration occurs only in boundary environments, it cannot be a correlate of a metrical system of alternating stress.

2.4 Summary

Labrador Inuttut is an Inuit language where Regressive Assimilation applies to all consonants clusters, making all underlying coda consonants part of a geminate in surface forms. Initial analysis ofthe acoustic results from the data considered here shows that SL behaves just as it is described in the literature, especially Dresher & Johns (1995). SL is exceptionless and independent of any metrical conditioning. No recurring pattern of prominence based on the three correlates of stress can be found. With these observations in mind, I address, in the next chapter, the previous literature on the SL phenomenon.

5 At the geographic language border with Yupik, the Lnuit dialect of Lnupiaq on Little Diomede Island in Alaska retains schwa.

27 3 Background Literature on SL, "stress" and pitch

3.1 Introduction

This chapter reviews the previous descriptions and analyses of SL, keeping the preliminary results in mind and trying to establish a theoretical baseline for a fuller examination of the results in Chapter Five. One ofthe goals in this thesis is to provide further empirical evidence documenting the SL phenomenon. Since no acoustic studies of

Labrador Inuttut are available in the previous literature, the theoretical basis for the interpretation of "stress" and pitch in this dialect primarily comes from acoustic studies of the related Inuit languages in Greenland and Quebec.

3.2 SL descriptions

Building on the basic description of SL in §2.1.2, I present in this section a chronological summary of SL descriptions in the literature. Nothing resembling the phenomenon is mentioned in the earliest literature on Labrador Inuttut. A diachronic study on the origins

SL is beyond the aims of this study. However the rule, on first glance, appears to be absent from 19th century lexicography and grammar books on Labrador Inuttut. Compare, for example, the following lexical entries from Erdmann ( 1864) and Bourquin ( 1891) to the same word in a more recent dictionary by Anderson, Kalleo & Watts (2006):

(1) 19th century orthography Contemporary Labrador IPA a. aksalloak 'wheel' atsaluak [atsaluak] b. Ketterdlermik 'jewelry ring' Kititlimik [xititiimik]

28 The examples in (1) suggest that SL did not apply previously, but more study is needed on the phonological assumptions of Erdmann and Bourquin, which are not described. It appears unlikely however that Bourquin would have overlooked SL if it was present in historic Labrador Inuttut, given his careful documentation of this dialect's polysynthetic morphology. One possibility is that they were influenced by the orthographic conventions of West Greenlandic, where SL is not a factor. The issue will not be resolved here.

3.2.1 SL in Quebec Inuttitut

SL gets its name from a French priest, Lucien Schneider, who spent most of his career working in the Inuit communities ofNorthem Quebec. Schneider (1966) wrote a series of grammar books where fa Loi des double consonnes (the law of double consonants) is first described. There is no analysis of the rule, but its output is consistent with the spelling in his 1970 dictionary, one of the most comprehensive lexicographies of any Inuit language.

It provides the data for the first formal description of SL as a phonological rule. Collis

(1970:276-77) dubs the phenomenon "Loi Schneider d'alliteration" (Schneider's Law of alliteration), shown in the following examples:

(2) SL description in Collis ( 1970): / aik + pa + SJi + aq/ 'fiance' West Greenlandic Quebec Inuttitut [a:ppaSJiaq] [aippasaq]

According to Collis, SL is found in all the Canadian Inuit dialects east of Hudson's Bay, including Quebec Inuttitut (which includes sub-dialects on the Belcher Islands

29 (Qikirtamiut), the eastern shore of Hudson's Bay (Itivimiut) as well as around Ungava

Bay (Tarramiutut)) and Labrador Inuttut (which includes sub-dialects spoken on the

1 North Coast of Labrador and in Rigolet). Collis argues that SL was operative in 19 h century Labrador, based on letters between Bourquin in Labrador and Kleinschmidt in

Greenland. He writes (1970:277): "cette regie existait deja au siecle dernier, ce qui explique pourquoi T Bourquin n 'a pu suivre les conseils orthographiques que S.

Kleinschmidt lui prodiguait dans ses lettres d 'aout 1865 et de juin 1871" (this rule existed in the last century, which is why T. Bourquin could not follow the orthographic advice that S. Kleinschmidt gave to him in his letters of August 1865 and June 1871). This assessment is not quite correct, as Bourquin did follow Kleinschmidt's orthographic advice in most areas, but this issue lies beyond the purposes of this thesis.

Following Collis, Rischel (1974:86) is the first to refer to SL as a rule governing the sequencing of syllables: "some Canadian dialects do not, according to the law, tolerate successions of closed syllables and hence a sequence VCCVCCV is simplified to

VCCVCV."

Massenet (1986: 131) shows that stating the rule in this way is too restrictive because simplification does not occur when a word's final syllable is eve, "si l 'on adopte cette interpretation, il faudra restreindre Ia regie aux syllabes internes de mot, puisque a Ia finale on peut avoir deux syllabes fermees successives: / sinippuq/ 'il dort'

/ ipp'it/ 'toi "' (if one adopts this interpretation, it [must] restrict the rule to word internal syllables, since at the end of a word there can be two successive closed syllables:

/ sinippuq/ 's/he sleeps' / ippit/ 'you'). Massenet takes instead a rule-based, generative

30 approach in his SL description. He also makes a convincing argument that in Quebec

Inuttitut the phenomenon operates iteratively, from left to right. Consider the underlying consonant clusters in the following data:

(3) SL description in Massenet (1986:125) a. / tussiaq-puq/ ~ [tutsiapuq] ' s/he prays' b. / tussiaq-vik-mut-lu/ ~ [tutsiavimmulu] ' and to church'

In (3a), the uvular is deleted while the labial is preserved at the surface level. The same happens in (3b), where, going from left to right, the final of the base is deleted while the final alveolar consonant of the next affix is preserved. For Massenet this is evidence that SL works from the left, targeting the first member of a heterogeneous consonant cluster. As support for this hypothesis, Massenet refers to a similar phenomenon in Labrador Inuttut described by Smith (1977) (see earlier in §2.1 .2) and in

Willis (1971:81), who argues the phenomenon is iterative and exceptionless: "in the

Ungava dialect there cannot be a sequence oftwo consonant clusters (or two tense consonants). When this occurs, through affixation mainly, the first consonant of the second (fourth, sixth, etc.) cluster is deleted."

Massenet (1986) contributes to the description of SL by showing where the rule occurs relative to other phonological rules in Quebec Inuttitut, especially in relation to a phenomenon that deletes the final consonant of base-stems. He calls this phenomenon effacement de Ia consonne finale du radical (called Truncation here, following Dresher &

31 Johns (1995)). First, Massenet demonstrates an environment where SL applies exclusively:

(4) SL alone (Massenet 1986:127) C,C2V(:)C3C4 Underlying / illu + rru:ja:q + tuq/ Truncation SL illuru:ja:qtuq Surface [illuru:ja:trtuq] maison-ressembledt-3s 'it looks like a house'

Massenet argues that in the case of a base-stem that ends in a vowel, the environment for

Truncation is not met, leaving SL to operate as in (4) where it targets Cdor deletion.

Massenet next illustrates environments where neither rule applies, then another where SL applies and Truncation does not, and finally two examples where Truncation applies and

SL does not. Consider the following:

32 (5) SL and Truncation (Massenet 1986:127) a. CN(:)C2C3 Underlying /niuviq + vik/ Truncation SL Surface [niuviprp'ik] faire.du.commerce-endroit ' store'

b. C,C2V(:)C3C4 Underlying / tussiaq + vik/ Truncation SL tutsiavik Surface [tutsiavik] prier-endroit 'church'

c. C,C2V(:)C3C4 Underlying /niuviprp'ik + liaq + puq/ Truncation niuviprp'iliaprpuq SL Surface [niuviprp'iliaprpuq] faire.du.commerce-endroit-aller- 3 s ' s/he goes to the store'

d. CC2V(:)C3C4 Underlying / qalluna:q + liaq + puq/ Truncation qallunaliaprpuq SL Surface [qalluna :liaprpuq] blanc-aller 's/he goes to the white person'

For Massenet, the crucial distinction is between (5b), in which the / -vik/ is

[-Truncation] and SL applies to the underlying heterogeneous consonant cluster, and ( 5c ), in which the suffix / -liaq-/ is [+Truncation] so that Truncation applies and SL is vacuous.

33 In his fmal case, Massenet shows environments where both rules apply, showing that

Truncation must apply first:

(6) SL and truncation (Massenet 1986: 128) a. CN(:) C2C3C4 Underlying tusaq-qqaujuq Truncation tusaqqaujuq SL [tusaqqaujuq] 'entendre-il-vient.de/just hear it'

b. c1c2 V(:) c 3c 4c s Underlying tussiaq-qqaujuq Truncation tussiaqqaujuq SL tussiaqaujuq Surface [tutsiaqaujuq] 'prier-il-vient.de/s/he is just praying'

c. CN(:)C2C3 Underlying ipa-ttauq Truncation SL Surface [ipattauq] 'veine.d'arbre-aussi/also a tree vein'

d. C1C2 V(:) C3-C4Cs Underlying ippaq-ttauq Truncation ippattauq SL ippatauq Surface [ippatauq] ' restes.de.nourriture-aussi/also leftover food'

Massenet's system of ordered rules is then expanded to include Regressive Assimilation.

Because Regressive Assimilation is crucial to his analysis of SL, discussion of that topic will resume in §3.3, which describes SL analyses in the scientific literature.

34 3.2.2 SL in the Inuit languages

Dorais ( 1976) calls "Schneider's Law of alliteration" a morphophonological characteristic of Quebec Inuttitut and Labrador Inuttut, describing it as an exceptionless law of elision.

He shows that SL does not apply in the Inuit languages west and north of Hudson's Bay by comparing lexical words there to the same words used by Inuit in Quebec and

Labrador:

(7) SL application, non-application in other Canadian Inuit languages (Dorais 1976:391) SE Baffin, Kinngaqmiut, Iglulingmiut Quebec Inuttitut/Labrador lnuttut a. autlaqpuq 's/he goes away' autlapuq b. akyakka 'my hands' akyaka c. ijukkaqtit 's/he makes him fall' ijukkatit d. utnukkut 'during the evening' utnukut

Dorais says SL affects surface forms in Quebec and Labrador so that in (7a) the second cluster / qp/ must be simplified because it follows / tl/ . Dorais differs with Collis in his diachronic assessment of SL. He argues that old Labrador texts and the memory of elderly Quebec Inuttitut language consultants show that SL is a recent innovation, coming into use sometime in the early 20'11 century. His main contribution relevant to the key aims of this thesis is his description of a dialectal continuum based on four variations: the voiced velar lateral approximant phoneme /1/, cluster assimilation, glottal stopping and

SL. His findings are summarized in the following table:

35 (8) Phonological processes in four lnuktitut dialects (Dorais 1976:391) lglulingmiut SE Baffin Kinngaqmiut Qubec Inuttitut Inuttut Itivimiut Taqramiut Presence of assimilation L --->t --->s --->s --->s L (-) Degree of cluster 0 1 2 2 3 neutralization 1 Presence(+) or absence(-) of + glottal stop Presence(+) or absence(-) of + + + SL

Only Dorais' descriptions of the phoneme /L/ in Labrador Inuttut is problematic: it is not supported by the data here or by the subsequent literature, including Dorais (1990b ). The table is otherwise useful, showing how dialects with SL are at the extreme end of heterogeneous consonant cluster assimilation.

Finally, Dorais describes what he calls a 'limited form' of SL in Northwest

Territories Siglitun (1986:46): "the first consonant of a cluster is elided in the same circumstances as described above, but only when the two elements of the group have the same position of articulation." Dorais provides the following examples:

The degree of neutralization varies from 0 in lglulingmiut (where all four types of clusters are fully used) to 3 in Labrador (where there is only one principal type).

36 (9) SL in Siglitun (Dorais & Lowe 1982:131) Siglitun Copper Inuit • a. iyluka 'my two houses' iylukka b. tikillijuiJ 'may he/she arrive at it!' tikilli d.fuiJ c. iylutka 'my (many) houses' iylutka

Dorais argues that SL applies in (9a-b) because the clusters /kk/ and / cH/ 2 have the same

' place of articulation, while SL has no impact on a coronal-velar cluster like / tk/ in (9c).

The primary goal here is to instantiate SL empirically. Lacking Northwest Territories I Siglitun primary source data to analyze, evidence from this Inuit language will play no further role in this thesis.

3.2.3 SL in Labrador Inuttut

Smith (1975: 105) writes that, in Labrador Inuttut, "two consonant clusters may not occur with only a vowel or vowel cluster between them, but must also have an intervening intervocalic simple consonant. There are no sequences of the form ... CCV(:)CC." No analysis is given, but Smith (1975:100) " the confirmability" ofthe hypothesis in Collis ( 1970:276-7) that SL existed in 19tl' century Labrador Inuttut, pointing to the

"unreliability of the orthography" and numerous counter-examples in Bourquin. Smith discounts for Labrador Inuttut the position of Rischel ( 197 4) that SL is a rule governing the sequencing of syllables (see §3 .2.1), showing that adjacent eve syllables are possible in / imappik/ ' sea' and / xaittuk/ 'band of land' .

2 Dorais describes / -3/ as a voiced glide in Copper Inuit, the apical fricative // in other Inuit dialects.

37 A more comprehensive description of SL is in Smith (1977b ), where he shows the behaviour of /-kkut/, the affix used in §2.1.2 to demonstrate SL in Quebec Inuttitut.

Smith finds the same pattern in Labrador:

(1 0) SL application in Labrador Inuttut (1977b:6) a. / nuna + kkut/ - [nunakkut] land. vial is

b. / tuttu(k) + kkut/ [tuttukut] caribou. vial is

Smith ( 1977b) also shows more complex constructions as evidence that SL applies iteratively from left-to-right:

( 11) SL application over longer sequences a. / nanu + IJIJUa(k) + xxa: + lluni/ [nanUIJIJuaxa:lluni] bear.toy.do first.by

b. / tuttu(k) + IJIJUa(k) + xxa: + lluni/ [tuttuiJuaqxa:luni] caribou.toy.do first.by

The affixation in both examples results in three adjacent underlying geminates, but the surface output in each case is different. The source of this variation must be a contrast in the base-stems. In ( 11 a), the base-stem has only short consonants, but in ( 11 b) the base- stem contains an underlying geminate in the suffix-adjacent position. Smith (1977b) therefore shows that SL must apply left-to-right, since only that direction can yield the

38 correct results in ( 11 ). The same pattern will be instantiated in the analysis here of the phonemic pair / tutuk/ versus /tuttuk/ (see methodology in §4.4.1, results in §5.1).

Smith (1977a) shows that, consistent with the results in Chapter Five, an open syllable interrupts the iterative pattern of SL, as we can see from the rule's non- application in the following example:

(12) SL blocked by the insertion of a CV syllable (1977a:82) Underlying / ximmi(k) + xa + IJIJik + tuk/ dog-have-3.s.neg Surface [ximmixaiJIJituk]

Smith argues that in ( 12) the underlying geminate / IJIJ/ does not de geminate because of the intervening open syllable / xa/ . This position is consistent with the preliminary results and will be further substantiated by the acoustic results in §5.1.

3.2.4 Deleting versus adjoining affixes in Labrador Inuttut

One of the issues not fully explained thus far is a variation first described by Smith

(1977a:8) involving two types of affixes differentiated by the way in which they adjoin base-stems with final coda consonants. One class, which he calls 'deleting affixes', is exemplified by / xa/ ' have' in (12). In that example the final stop in the base-stem

/ ximmi(k)/ is deleted, not because of SL, but because / xa/ is a deleting affix. The other class of affixes preserve the final coda consonant of a base-stem they adjoin, which Smith

39 calls 'adjoining affixes' (Massenet (1986) describes the same phenomenon in Quebec

Inuttitut as [±Truncation]). Both classes are shown in the following data:

(13) Deleting versus adjoining affixes in Labrador Inuttut Smith (1977a:8) a. /inuk + IJa/ -+ [inuiJa] person.3poss

b. /inuk+mit/ -+ inukmit [inummit] person. from

The deleting class is exemplified by /IJa/ in ( 13a), while / mit/ is an example of an adjoining affix (note the preservation of the final stop is here made opaque at the surface level by the application of Regressive Assimilation to the underlying heterogeneous consonant cluster /km/). Smith acknowledges that this alternation is not the same for all language consultants:

... the Labrador dialect has been in a state of rapid change for at least the last century. The adjoining/deleting classes are presently quite variable from idiolect to idiolect. (Smith 1978:116)

Smith's dictionary of affixes (1978) is the basis for most of the glosses used in this thesis, but he admits that for some cases, "insufficient or contradictory data was obtained.

Individual speakers may in certain cases exhibit the class which is not given" (p. 116).

These observations of variability in Labrador Inuttut are consistent with the findings in

§5.1.4 where SL is shown to be exceptionless. At the same time, the perfect alternating pattern shown in (11), where SL applies to all syllables, is not always the case in the data considered here which show some variability at the right edge of words for the affix

40 [IJI)ituk], [I]ittuk], or the previously unattested [IJituk] . Crucially these variations never cause an SL violation.

3.3 SL analyses

Only two studies of SL attempt a principled explanation and theoretical solution to the phenomenon. Massenet ( 1986) argues for an articulatory motivation related to "tense" geminates. Dresher & Johns (1995) offer Government Theory as a plausible framework of analysis.

3.3.1 SL is not "geminate tension" in Quebec Inuttitut

Massenet (1986) shows that in Quebec Inuttitut Regressive Assimilation is not total, allowing a class of clusters that begin with a uvular [q x If N] such as / qp/ and / lft/ . From these uvular clusters, Massenet finds phonological processes that result in pharyngealized geminates which he calls r-clusters [prp], [prp'], [trt], [trs], [trs'], [1"1], [mrm], [nrn] and a third class of consonant clusters he calls glottalized geminates [pp'], [ts'], [kk'] and [qq'].

All other consonant clusters are surface geminates with SL degeminating each as follows:

[vv ~v], [pp ~p], [ts'~ j], [11 ~1], [tt~t] , [kk~k] , [kk'~y], [qq~q] and [qq ' ~lf]. He argues that in Quebec Inuttitut both geminates and uvular clusters are tense as compared to the unassimilated clusters in other Inuit languages. According to Massenet (1986: 130), the idea that geminate tension is the force behind SL follows Schneider's (1970:XIV) insight that geminates are "tendues," a term that Massenet interprets to mean articulatory tension. He argues for Tension (TEN) within the list of ordered rules already mentioned:

41 Truncation, Regressive Assimilation, SL, Affrication and Devoicing. Dresher & Johns

( 1995 :85) describe Massenet's argument as articulatory, with the production of consonant clusters causing an "explosive release of air" or consonant tension. To avoid a 'tension clash' in adjacent syllables, Massenet (1986:105) proposes the following rule:

(14) SL: Law of double consonants (La Loi des double consonnes) RULE: Delete a word-internal coda consonant in a syllable with a tense onset

Massenet argues for the ordering of his phonological rules as (Regressive Assimilation~

Tension~ SL) applied in derivational cycles, to yield the correct surface forms as follows:

42 (15) Rule ordering in Quebec Inuttitut (Massenet 1986: 131 -2) a. /niuviq+vik+mut/ 'to the store' 1er cycle: niuviq (aucune regie ne s 'applique) 2eme cycle: niuviqvik R. Assimilation niuvivrvik Tension niuviprp'ik SL niuviprp'i 3eme cycle: niuviprp'imut

[niuviprp'imut]

b. /anaH+Haj + jaiJ + IJit+ tuq/ 'slhe didn't return' 1er cycle: allaH

2eme cycle: anaHHaj R. Assimilation Tension anaqq'aj SL anaqq'a 3eme cycle: anaqq'aja

4eme cycle: anaqq'ajaiJIJit R. Assimilation Tension anaqq'ajaiJit SL anaqq'ajaiJi 5eme cycle: anaqq'ajaiJIJituq

[anaqq'ajaiJIJituq]

Tension occurs in the second cycle of ( 15a) where its output is a glottalized geminate,

Massenet's SL trigger environment. In the second cycle of ( 15b) , Massenet observes the same process for the uvular stop. What he does not explain is the fourth cycle where

Tension appears to delete the / IJ/ segment. This is either a printing error or Massenet envisaged an unstated definition for tense onset in (14) that includes the nasal geminate

/ IJIJ/. His stated definition of tense consonant is a 'delayed burst', which cannot apply to

43 non-. While agreeing with Massenet's account of Regressive Assimilation, I find

no motivation for Tension as an articulatory phenomenon, the basis of which also makes

the wrong predictions with respect to the observed facts of SL in Northwest Territories

3.3.2 SL is not a metrical phenomenon

Dresher & Johns (1995) is perhaps the most ambitious study of SL in the literature. The

authors first use feature geometry to explain the degree of assimilation in the three dialects of lnuktitut where SL is attested, then they show how this phenomenon cannot be

metrical or related to a compensatory phenomenon observed in some Inuit languages, and conclude with an overall theoretical solution based on Government Phonology. They propose:

3 As discussed in § 1.1 , SL is reminiscent of other dissimilation processes, especially those that impact length. The rule in Northwest Territories Siglitun also looks like a dissimilation process involving the nasal feature in Gooniyandi. McGregor ( 1990:98) shows the progressive dissimilation ofNC clusters, saying that the rule "deletes the nasal in a homorganic nasal-stop cluster when it immediately follows any nasal-stop cluster." Consider the behaviour of the ergative postposition / -IJga/ which loses its initial nasal when the preceding syllable boundary is also a nasal-stop cluster, as seen in the following examples: / go: Il.bo:-l)ga/ [go: 11.bo:ga] ' by the woman' / gongo:do: Il.bo:-l)ga/ [gongo:do: Il.bO:IJga] ' by the snot' McGregor further shows that the Gooniyandi dissimilation rule avoids "homorganic nasal-stop clusters [which] follow one another in successive syllables [ ... and] sequences of non-homorganic nasal-stops followed in the next syllable by homorganic nasal-stop clusters." By comparison, an SL trigger in Northwest Territories Siglitun can be a cluster (but also a geminate) and the rule only simplifies geminates: clusters cannot be SL targets, as shown in §2.2. The crucial difference is that SL is not morphologically conditioned; Suzuki(1998: 155) shows that the NC cluster dissimilation in Gooniyandi operates within the morphological domain . See further discussion of dissimilation processes in §5.1.4.

44 (16) SL Description: VCCV(:)C 1C2V- VCCV(:) C2V SL Rule: Delete the left root node of a place geminate when it follows a consonant cluster

As discussed in § 1.3, Creider (1981) observes a typological difference in Regressive

Assimilation across the Inuit languages. Dresher & Johns (1995:86) describe the situation in terms of feature geometry. In Northwest Territories Siglitun, place is not assimilated, but voicing, nasality and continuancy are. They propose Manner as one feature node and split the Place node based on the facts of all three dialects:

(17) Continuum of Regressive Assimilation (Dresher & Johns 1995:86) Siglitun: Manner Quebec Inuttitut: Manner Primary Place Labrador Inuttut: Manner Primary Place Secondary Place

Dresher & Johns (1995) on the exception to place assimilation in Quebec Inuttitut, described above as 'r-clusters.' Massenet (1986) shows that these pharyngealized coronal

1 1 or labial geminates [prp], [prp ], [t't], [trs], [trs ], [1'1], [mrm] and [nrn] lose their pharyngealization when targeted by SL, shown by the following data:

(18) SL and ' R-clusters' in Quebec Inuttitut (Dresher & Johns 1995:87) Underlying I aullar + tul)a/ SL aullatul)a Surface [aullatul)a] (*aullartul)a)

45 Since Regressive Assimilation does not "wipe out the pharyngeal element contributed by

/r/," Dresher & Johns (1995:87) argue that the data in (18) means that all the other features must spread together. They give pharyngealization an independent status, consistent with the fact that it is a secondary articulation:

(19) Basic Feature Geometry Pharyngeal SECONDARY PLACE 0 I ROOT NODE 0 I PRIMARY NODE 0 ------~ PRIMARY PLACE ~---- Labial Dorsal / Cor~nal MANNER 0 Cont Voice Nasal

From (19), Dresher & Johns (1995) can say that Regressive Assimilation in Northwest

Territories Siglitun spreads the manner node, while in Quebec Inuttitut it spreads the primary node. Consistent with this architecture is the fact that, unlike other features, pharyngealization also spreads to vowels. Finally, in Labrador Inuttut, which lacks pharyngealization, Regressive Assimilation spreads to all nodes. This is consistent with the preliminary results and the results in Chapter Five which show that Regressive

Assimilation in Labrador Inuttut is total. 4

4 One exception are those examples from the Rigolet sub-dialect, in which heterogeneous consonant clusters not heard in other parts of Labrador are valid. Initial results from HP, the Rigolet Inuktut consultant, show that the phenomenon affects the coronal place of articulation: [yl], [{31] , [x t], etcetera (see also Dorais ( 1977b ), Dresher & Johns ( 1995), and Bobaljik ( 1996) for related discussion, especially the Dresher & Johns ( 1995:93) discussion of the two environments where SL remains

46 One of the outcomes of this thesis is that SL is not a metrical rule, in line with

Dresher & Johns (1995). They consider three plausible metrical solutions and then show

how each one fails to explain SL. The first proposal supposes a system of strong and

weak syllables from left to right, with SL applying to syllables in the weak position as

follows:

(20) Metrical Theory One: S syllables left-to-right, W undergo SL a. bear. toy. do-first. by 'by first killing the toy bear' 2 3 4 I 2 3 4

/ nanu + IJIJuak + kxa: + lluni/ ---+ [nanUIJIJUaxa:lluni] S W S* W S W b. caribou. toy. do-first. by ' by first killing the toy caribou' I 2 3 4 I 2 3 4

/tuttu + IJIJuak + kxa: + lluni/ ---+ [tuttuiJuaqxa:luni] s w s w s w

This analysis accounts for the pattern in (20b ); however SL is sensitive only to the preceding syllable. As a result, in (20a), SL simplifies the underlying geminate in consonant position three, therefore weakening a syllable that is supposed to be strong.

inactive: in words with only CV sequences and in words with a string of mixed clusters dissimilar in their place of articulation. Dresher & Johns note that later environment could potentially occur in Rigolet Inuktut, as it does in Northwest Territories Siglitun. In their footnote on page 93, Dresher & Johns mention a forthcoming paper, but state that their "results, so far are inconclusive." Results from the Rigolet Inuktut consultant are the same: while numerous examples of mixed place of articulation clusters do arise, in only one case from several hours of recordings is there an example of syllable adjacent mixed clusters and this is a marginal case at best. Further analysis is not possible in this thesis. Results from the Rigolet lnuktut consultant do confirm that Regressive Assimilation does not spread to coronal node, suggesting a different feature geometry for the Rigolet dialect is needed. This as well is a topic for future study.

47 Dresher & Johns (1995) also show that reversing the strong weak pattern does not resolve the problem:

(21) Metrical Theory Two: W S syllables left-to-right, W undergo SL a. bear. toy. do-first.by 'by first killing the toy bear' I 2 3 4 2 3 4

/ nanu + IJIJUak + kxa: + lluni/ ---+ [nanUIJIJuaxa:lluni] ws w s ws b. caribou. toy. do-first.by 'by first killing the toy caribou' I 2 3 4 12 34

/ tuttu + IJIJUak + kxa: + lluni/ ---+ [tuttuiJuaqxa:luni] W S W S* WS

This account describes (21a). But in (21b), Dresher & Johns (1995) observe that SL degeminates the underlying geminate at consonant position four, thus simplifying what is supposed to be a strong syllable. They further note that any system of strong and weak syllables that one could imagine fails each time an open syllable is inserted in the relevant string. Dresher & Johns (1995 :89) conclude that a metrical explanation "requires that the metrical system locates heavy syllables wherever they are." They next consider SL as a stress-governed rule, assuming that closed syllables are stressed and that adjacent stressed syllables clash. SL would thus resolve the clash by deleting the coda of the rightmost closed syllable. The proposal is schematized in the following rhythmic grid:

48 (22) Metrical Theory Three: Assign stress to every closed syllable, SL context is adja~_en! h.:~avy syllabl~s ~ * * * * * * * * * * (* *) (* *) Jl Jl Jl Jl Jl Jl Jl Jl Jl Jl Jl Jl Jl Jl eve eve eve eve - eve ev eve ev

Under this system coda consonants add one mora to the overall weight of any syllable.

This works as long as the vowel adds only one mora to the grid. But consider the same system for the examples in (20-21) where some syllable peaks are bimoraic:

(23) Metrical Theory Three does not work for long vowels a. --. * * (* *) (* *) (* *) ' ! fl flfl flfl flflfl fl fl ev eve evv evve ev ev na llUIJ IJUa xa:l lu ni b. i * * . (* *) (* *) (* *) ' flfl fl flflfl flfl fl fl I eve ev evve evv ev ev 1 tut tu IJUak xa: lu ni j

The problem, in Dresher & Johns's (1995) view, is that long and/or complex vowels cannot be metrically weak, the situation for the third syllable, [IJua], in (23a) and the antepenultimate syllable [xa:] in (23b). Finding no metrical solution to the problem,

Dresher & Johns (1995 :90) next consider a diachronic proposal from Ulving (1953). He argues that "Inuit has, or once had, a rule of " so that an original /k/

49 is weakened to /y/ when it precedes main stress, as in the word [putuyuq] 'big toe'.

Ulving (1953) argues that !k/ is preserved and geminated when it follows main stress as in the following examples where stress shifts back a syllable in the singular versus dual forms:

(24) Consonant Gradation (Ulving 1953, adapted from Dresher & Johns 1995:90) a. nukaq --+ nukkat 'younger sibling' 'younger siblings (2)' b. pu:q ( < *puyuq) --+ puyyut ' bag' ' bags (2)'

Rischel (1974) agrees that consonant gradation may previously have been a factor but he rejects a stress-based solution for modem day Inuit languages, especially West

Greenlandic where he argues that 'stress' or intensity prominence is undefined. According to Rischel, phenomena like those found in (24) are related to compensatory lengthening in West Greenlandic. The rule is like SL in that it also involves geminates alternating with non-geminates but is otherwise unlike SL, being morphologically conditioned, sensitive to vowel length and non-iterative. Synchronic evidence that SL and compensatory lengthening are unrelated comes from Northwest Territories Siglitun where both rules operate independently, as shown in Dresher & Johns (1995:92):

50 (25) Gemination and SL in Northwest Territories Siglitun (adapted from Lowe 1985) a. b. c. Singular iyaliq quaq upkuaq Dual Formation iyallak qujjak upkuajjak SL upkujak Affrication qud3ak iyallak qud3ak upkujak Gloss = dual of: ' window' 'frozen meat' 'door'

d. e. f. Singular iqidjralik itiyaq sulukpauyaq Dual Formation iqidjrallak itikkaq sulukpaukkaq SL iqidjralak sulukpaukaq Affrication iqidjralak itikkaq sulukpaukaq Gloss = dual of: 'square' 'foot' 'grayling'

Dresher & Johns ( 1995) argue that since SL operates on the output of gradation (or compensatory lengthening) as in (25c, d, f), the rule must be independent from any gradation phenomenon.

Dresher & Johns ( 1995) finally consider consonant gradation in Inupiaq, spoken on the Seward Peninsula of Alaska. This Inuit language is characterized by Kaplan (1985) as having a pattern of geminationldegemination that regulates an alternating pattern of strong and weak syllables. According to Dresher & Johns (1995 :92) the pattern is set in motion by the first syllable of the base-stem: "if it is closed or has a long vowel the pattern begins with a strong syllable; if it is open with a short vowel the pattern starts with a weak syllable." Consider the following data:

51 (26) Seward Peninsula Consonant Gradation (Kaplan 1985) a. /tuttuttuq/ - [tuttutuq] caribou.kill.3s b. /katittuq/ [katittuq] marry.3s

Dresher & Johns (1995) acknowledge that the rule is iterative like SL, outputting C-CC or CC-C sequences. The similarities, however, end there. Consonant Gradation is sensitive to vowel length and functions as a rhythm rule, maintaining iambic stress.

Dresher & Johns (1995:93) argue that SL cannot have a rhythmic function since it remains inactive in CVCV and CiqV(:)C;CiV(:) sequences: "it is only when a cluster is followed by a geminate that SL is brought into play." This singular generalization is fully consistent with the durational pattern observed in the primary data considered in this thesis.

Dresher & Johns ( 1995) conclude with a proposal whereby the source of SL lies in a government relationship between coda consonants and the following onsets (PN = place node):

52 (27) SL and Syllable Government

Sl S2 S3 S4

0 R 0 R 0 R 0 R

N N c N c N

Cl v v C2V v C3 C4 v C5 C6 v

PNl PN2 PN3 PN4 PN5

The first syllable has no coda and thus no relationship with the following syllable. But in the second syllable, Dresher & Johns (1995) argue that the coda is governed by the onset that follows it. Since a governing syllable cannot itself be governed, the motivation for

SL is to eliminate that possibility by deleting C5 and thus any trans-syllable relationship between S3 and S4. While the proposal accounts for the data, the nature of government, or of the factors that ultimately drive government relations, remains undefined (Rose,

Pigott & Wharram 2012).

3.4 Accounts of "stress"

The nature of syllable prominence in the Inuit languages is unclear in the literature. 5 The issue is discussed in some detail in a recent acoustic study (Jacobsen 2000), which concludes that West Greenlandic is a not a stress language. The results presented in

Chapter Five suggest that this is also the case for Labrador Inuttut.

5 Except for Seward Peninsula lnupiaq which, according to Kaplan ( 1985), has iambic stress.

53 3.4.1 Early accounts of "stress" in West Greenlandic

In his orthography of West Greenlandic Egede (1760) uses long and short accents to represent phonemic length, but he leaves the nature of stress undefined. Nowhere in the

Inuit languages literature is there a convincing description of the correlates of prominence. Rischel (1974:91) writes that "the category of stress has no well defined status in West Greenlandic phonology." Unlike intonation, he says, it is difficult to make generalizations about stress patterns, though the tendency is to perceive stress on the final syllable and/or the antepenultimate. Kleinschmidt's (1851 :8) acoustic impression of West

Greenlandic is that stress generally falls on the penultimate syllable, with stress called

"ton" (accent) and defined as a "hebung der stimme," which Jacobsen (2000:41) translates as a "raising of the voice." However the acoustic manifestation of stress is left undefined. Kleinschmidt later presents an account of stress based on syllable weight: a coda consonant counts 1, a short vowel 2, and a long vowel 4. This configuration supposedly yields the correct word stress in the following phrase:

(28) Kleinschmidt: stress based on syllable weight 2 2 22 3 23 53 23 3 4 nanu ilumut qilammiituq itsirssarppaa

Kleinschmidt argues that syllable with the highest count in (28) get the most stress, with lesser amounts of stress going to each syllable with lower counts. According to his account, each word has one main accent and one subsidiary accent falling on the initial or ultimate syllable, while long words may have several subsidiary accents. The distribution

54 is as follows: heavy syllables always attract accent; two adjacent syllables cannot both have main accent; there cannot be more than two syllables in a row without accent; and main accent falls heaviest on the last three syllables unless they are equal in which case stress falls on the third syllable from the end. Despite the elaborate system described in

Kleinschmidt, Jacobsen's (2000) study of stress correlates in modem West Greenlandic reading data shows that stress has no reliable acoustic basis. This finding is consistent with the data considered here. Despite these acoustic realities, the Inuit languages are often described in the literature as having a system of stress that is sensitive to syllable weight.

Rischel (1974:78-80) contends that the undefined 'accent' in Kleinschmidt may be pitch. His book on West Greenlandic phonology includes a chapter on phenomena related to syllabification. Addressing the representation of long segments, he shows a prominence pattern for the stress correlate of pitch: "for example the phrase final neutral intonation contour high-low-high" (p. 78). Rischel explains the timing of this HLH boundary melody in terms of morae: each short vowel in West Greenlandic can carry one tone. In other words, it has one mora; each long vowel has two morae and can thus host two tones and so on. Consider the following data:

55 (29) Rischel: vowel morae explain the timing of boundary melodies in West Greenlandic

a. HLH HLH HLH I I I I II Il l akivara aavaa ataasiq 'I answer him/her' 's/he fetched it' 'one'

b. HLH HLH HLH I I I I I I I II uvaiJa uvaiJalu uvaiJattaaq 'I' 'and I' '1, too'

c. HLH HLH HLH I I I I II II I akivat akivaa akivaatit 'you answered him/her' 's/he answered him/her' 's/he answered you'

d. HLH HLH HLH Il l I II Il l ataasiiq tiguwaa tiguwaaa 'one?' 's/he takes that' 'does s/he take that?'

e. HLH HLH I l l Ill apirai apiraai 's/he asked them' 'did s/he ask them?'

The data in (29) indicate that the HLH boundary melody is timed with the three rightmost morae in any phrase. Those data also indicate that the interrogative can involve a lengthening of the final syllable, which then causes the HLH boundary to shift relative to the base-stem, as in (29a) [ata:siq] versus the question form in (29d) [ata:si:q]. This lengthening can also result in homogeneous overlong vowels as in (29d) and heterogeneous overlong vowels as in (29e). This thesis will not deal with the complex question of how to syllabify these sequences, but Rischel's insights on intonation will be

56 touched upon again in §3.4.3. In this section about "stress," Rischel's contribution is to show that this prosodic phenomenon of intonational phrasing may have been incorrectly perceived by Kleinschmidt and others as word stress.

One consistency with Kleinschmidt's account throughout the literature is the position that heavy syllables always attract stress in West Greenlandic. Dresher & Johns

(1995:89) write that while "the facts of stress in Inuktitut tend to be elusive, on most accounts syllables with long vowels or vowel clusters have some degree of stress: typically they have the most stress." The preliminary results here show that this is sometimes the case. But the full results in Chapter Five show that short vowels can also attract peak intensity and host pitch effects. These findings are also contrary to another system of stress proposed by Smith (1975: 103-4) for Labrador Inuttut. While he acknowledges that his account is "aural impressionistic and not based on spectrographic analysis," Smith nevertheless proposes the "gross features of stress assignment at the word level" as adhering to a ternary system where the strongest [1] stress falls on syllables with a long vowel, the next strongest [2] on precluster syllables and the least stress [3] on open syllables. Consider the following data:

57 (30) Three degrees of stress in Inuttut (Smith 1975:103) 23313133 nukx.akasa:ligi:kxuiJa 'I am already almost finished'

Like Kleinschmidt ( 1851 ), Smith does not define stress and offers no acoustic evidence of the correlates he may be referring to. As we will see in Chapter Five, however, no one correlate or any combination of correlates could be found to match the pattern in (30).

Intensity, for example appears to be entirely at the discretion of the speaker. Minimally,

Smith (1975: 103) contends that "subtler factors such as emphasis and emotion can be superimposed" onto his proposed system for word stress.

3.4.2 A recent account of "stress" in West Greenlandic

Jacobsen attempts to resolve the outstanding question of stress in West Greenlandic in her

(2000) study of durational and pitch values for a series of words in carrier sentences read seven times by two West consultants. Her focus on just two possible correlates of stress comes from Rischel (1974:96):

... two prosodic parameters must be studied thoroughly before it is advisable to speak of stress. One is intonation in relation to vowel morae. The other is Kleinschmidt's concept of syllable weight. Since the latter parameter is entirely deducible from the segmental structure of word forms it is no problem to represent it consistently, and hence it should be entirely possible to test empirically to what extent the subjective category of stress can be explained as a complex function of syllabification, syllable weight, and intonation. If there is a residue of unexplained rhythmicization (which there is, without doubt), we may begin to search for a significant parameter of stress' .

58 As mentioned above, Jacobsen (2000) finds no empirical basis for Kleinschmidt's system of stress based on syllable weight. Her working hypothesis is that, minimally, stress must include more than one acoustic parameter. Therefore the relevant parameters in her view, duration and pitch, "must covary in a systematic and consistent way." She finds no evidence of this covariance in her data and concludes that stress is not a "relevant category in the description of West Greenlandic word prosody." This finding is consistent with the detailed results presented in Chapter Five.

3.5 Accounts of pitch in the Inuit languages

The most studied phonological pattern in the Inuit languages involves intonation, introduced already for West Greenlandic in §3.4.1 as being moraically timed. I conclude this chapter with a review of the scientific literature on intonational phrases in West

Greenlandic and Quebec Inuttitut, which serve as background to the analysis of pitch patterns in Chapter Five.

3.5.1 Accounts of pitch in West Greenlandic

Mase & Rischel (1971:235), Mase (1973), and Rischel (1974) examine pitch in West

Greenlandic. Each concludes that the syllable is a functional category. Further, they argue that since "intonation is clearly based on a mora-counting principle, we have two units of measure in West Greenlandic: VOWEL MoRA and SYLLABLE" (p. 97). Rischel also describes five terminal contours (that will not be addressed in detail here), an important one being the phrase-final boundary melody HLH. Exhaustive FO studies of reading task

59 data by Nagana-Madsen (1988, 1990, 1993, 1994) and Nagana-Madsen & Bredvad-

Jensen (1995) reveal empirical evidence ofthese phenomena. Gussenhoven (2000:133) summarizes their findings, using examples from Rischel (1974), including (29d-e), which highlight the contrasting tonal patterns in declarative and interrogative phrases with the addition of a vowel mora to the final syllables of the examples in (29e ). Consider the following data:

(31) Intonation in West Greenlandic: declarative versus interrogative a. taku4iuk takuwfiu:k 'you saw him/her' 'did you see him?'

b. tsiguwa: tsiguwa:: 's/he takes that' 'did s/he take that?'

c. apirai apira:i ' s/he asked them' 'did s/he ask them?'

First from a segmental perspective, Gussenhoven observes that final short vowels in declaratives are lengthened in the interrogative form, as shown in (31a). Final long vowels in declaratives become overlong in the interrogative form, as shown in (31 b).

Similarly, final diphthongs in declaratives also become overlong, with the first element of the diphthong being lengthened, as shown in (31 c). Thus, from a mora count perspective, the final syllable in (31 a) is monomoraic in the declarative and bimoraic in the interrogative, while the final syllables in (31 b, c) are bimoraic in the declarative and trimoraic in the interrogative. Nagana-Madsen (1993) adds one more detail to Rischel's

(1974) account ofthe terminal HLH contour, decomposing that boundary melody into a

60 word-final boundary melody ofHL and a phrase or utterance-final H boundary tone. Her argument is thus that all words have an HL boundary melody, which is generally consistent with the current results, and that the H boundary tone is phrasal, falling on the final mora of a phrase or utterance and causing the HL boundary melody to shift leftwards by one mora. Finally, based on the results from two language consultants reading six paragraphs of West Greenlandic text, Nagano-Madsen & Bredvad-Jensen

( 1995) treat the HL previously described as a word domain boundary melody as an FO reset or, "a pitch accent which appears in relation to a word boundary, while the phrase­ final H tone can be referred to as phrase accent" (p. 152). Support for this analysis comes from the fact that in the phrase-final position they observe that with the reset of L, it "is nearly 100% predictable that it will be followed by a phrase-final H tone which is manifested as a FO rise." (Nagano-Madsen & Bredvad-Jensen 1995:137).

I assume, as a starting point, that the body of work on West Greenlandic is directly applicable to the related language of Labrador Inuttut studied here. It therefore stands as the theoretical model for intonation used in my analysis ofFO results in Chapter Five, where I show that boundary melodies in Labrador Inuttut also consistently pattern with various dialogue contexts.

3.5.2 A recent account of tonal and durational patterns in West Greenlandic

Building on her analysis of the prosody of West Greenlandic, Jacobsen makes three further conclusions, summarized as follows:

61 (1) The prosodic characteristics of words can be explained in either tonal or durational terms. (2) The four different syllable types (of different 'weight') are distinguished in durational terms; further, there appears to be only a tripartite system of short, long and overlong [segments]. (3) There are intra-syllabic as well as inter-syllabic rhythmical adjustments. It is concluded that Greenlandic prosody does not include an autonomous stress category, either tonal or durational parameters alone will do. And although Greenlandic has distinctive quantity, there is room for considerable durational variation of segments. (Jacobsen 2000:40)

The first point is consistent with the results of the current investigation, where, as we will see, intensity is unsystematic. The second point is also consistent with the results detailed in Chapter Five, which generally reveal four classes of syllable length and three classes of syllable peak length (no overlong consonants). Like Jacobsen's work, this thesis presents empirical evidence of distinctive quantity and also shows a limit to durational variation: phonemic long and phonemic short segments never overlap.6 Jacobsen's (2000) study

6 This study was not designed to test for the " intra-syllabic as well as inter-syllabic rhythmical adjustments" that Jacobsen finds in her data. She describes these, respectively, as (a) shortening oflong segments adjacent to other long segments, and (b) shortening of either the first or second syllable from adjacent heavy syllables, a so-called 'weight clash'. Dealing with the segmental adjustment first, Jacobsen observes:

[ ...]the difference (both relative and absolute) between phonologically long lkk/ after the long vowel in atu:kkasura: and the phonologically short /k/ after the long vowel in piku:kulavuyut happens to be smaller than the difference between the phonologically long /kk/ after the long vowel in atu:kkasura:) and phonologically long /kk/ after short vowel (in kukukkumavara). Jacobsen (2000:58)

Jacobsen argues that her West Greenlandic language consultant uses this durational variation as a form of rhythmicization, with the limitation that short vowel duration must not exceed the duration of its long vowel counterpart. This phenomenon will not be explored in this thesis. The second durational phenomenon Jacobsen (2000) argues for is that adjacent super heavy syllables CV:C create a weight clash that her language consultants resolve by shortening consonants in either the first or second syllable. The environment arises in the test word [ta:ma:lla:llia:si:t], where she observes that one of her language consultants reads the first CV:C more quickly than the second CV:C, while the other language consultant does the opposite. The preliminary results here show no evidence of this phenomenon and it will not be pursued in this thesis. However, these detailed analyses of segmental duration do inform the establishment of the acoustic criteria for geminate status, crucial in Chapter Five to the instantiation of SL in the data considered here.

62 presents empirical evidence, consistent with the findings ofNagano-Madsen (1992:62), that geminates in West Greenlandic are twice as long as their short counterparts.

3.5.3 Pitch patterns observed in Quebec Inuttitut

Finally the pitch effects observed in the current data are consistent with Massenet's

( 1980) findings in his acoustic study of spontaneous speech recorded from Inuit language consultants originally from Quebec, Canada. 7 His analysis of FO for these Quebec

Inuttitut speakers shows patterns for declarative, interrogative and imperative phrases. As regards final declarative intonation, Massenet observes high tone on the penultimate vowel, as shown in the following examples:

(32) Massenet: declarative intonational melody (1980: 197-8)

Hz

300

'j • a i mange '

7 Massenet interviewed speakers living for more than 20 years in Resolute Bay, Nunavut (now called Qausuittuq), but originally from Port Harrison, now called lnukjuak, in Quebec.

63 Hz

300

250

200 ~

uo ·-1 - f---f.- p 1 j a r if roo u Q 'U /e.st Hni'

cs 1-t 10 20 30 40 ~ o

Hz

300 - 250 - - 200 -

150 k a a t t ui t ' qul ont faim'

In the left and medial syllables of the first example in (32), pitch is flat or slightly falling from left-to-right across each syllable, a pattern shown in the Chapter Five results to be the unmarked pitch pattern. By contrast, at the right edge of the declarative phrases in

(32), Massenet's data show high pitch on the penult consistently contrasting with low pitch on the phrase-ultimate syllable. This finding is consistent with the declarative phrase data shown in Chapter Five. Massenet's description is also consistent with

64 arguments for an HL boundary melody associated with the final two morae of phrases in

8 West Greenlandic, as described in Gussenhoven (2000) •

Massenet (1980:204) next describes interrogative phrases where "!'accent musical se place sur l 'ante-penultieme" (boundary melody is placed on the antepenult). Massenet describes three different interrogative types. First, for questions where the answer is known or visible, Massenet finds no lengthening of the phrase-final syllable, as shown in the following results:

(33) Interrogative intonational melody, no lengthening (Massenet, 1980: 199-200) 30::i ::::; ~

150 ·~~----!!----t---t- ' t I I p i j a r fi a ' Jl a/e6t fin l ? '

Hz 300 -

2:SO -

200 - __/ 0 l50

I I r 0 i X i 1 iU op u I) a ' j ' ili mange?'

8 Nagano-Madsen & Bredvad-Jensen ( 1995) put forward an alternative analysis to the position that West Greenlandic words end with an HL boundary tone, arguing instead that the observed phenomenon may in fact be an FO reset.

65 Hz 300 -

25 0 - ~

200 -

150

k ~ a pp t ' as-tu faim ?'

Hz 300 - / 250 -

200 -

150 k a.S tt ui t ' qui ont falm ?'

Massenet observes that "!'accent musical est place sur La troisieme voyelle La fin de La phrase" (the boundary melody is placed on the third vowel from the end of the sentence), as shown in (33) where the H tone begins on the ante-penultimate syllable for each example.

In the second type of question, the answer is not known. Massenet ( 1980:200) describes lengthening for this type of interrogative phrase in a phenomenon called

"redoublement" (reduplication) of the phrase-final vowel. In these cases the high tone is again placed on the ante-penultimate syllable, as shown by the following data:

66 (34) Interrogative intonational melody, with reduplication (Massenet, 1980:200-201)

300 250 - --- 200

150

k aa tt tii f t 'qui on t faim ?'

11%1 300 -t I I _...... _ 250 -r- I 200 1 1)0 I I k a a DP ; 1 t ' as-tu faim ? '

Massenet argues for a different pitch pattern in these types of questions, since, following the H tone on the antepenultimate syllable, he observes that "apres la retombee sur la voyelle suivante, on assiste a un (legere) remontee sur la derniere" (after the decline of the following vowel, there is a (slight) rise on the last). The data in (34) are thus consistent with the descriptions of an HLH boundary melody in West Greenlandic, and with the pitch results in the data considered here.

67 The third type of question echoes a statement, where the questioner is looking for confirmation as a result of misunderstanding or surprise. Massenet (1980) observes

"surallongement" (overlong) in these cases, shown by the following examples:

(35) Interrogative intonational melody, with reduplication (Massenet, 1980:200-201)

Hz

:JOO -

250 -

150 I I l I I I r I I I n i X i l i u pp u I) aaa 'j ' ai mange ? '

Hz

300 -

250 -

200 -

150

k aa pp fii t ' as-t \l fa im ? '

Massenet argues for the same pitch pattern as (34), except that in (35) the HLH melody falls entirely on the overlong final vowel. This description is consistent with the accounts of West Greenlandic, and, further, the same pattern is observed in the results for pitch discussed in Chapter Five.

68 3.6 Summary

SL is a phenomenon in which the second of two syllable adjacent geminates is degeminated. The rule is described in the literature for Quebec Inuttitut, Labrador Inuttut and Northwest Territories Siglitun, where it is shown to exclusively target underlying geminates. It is indifferent to vowel length, and works on the output of other rules like

Truncation, Regressive Assimilation and Consonant Gradation. SL is not related to any metrical pattern. In regard to rhythm, studies on the Inuit languages show no evidence of metrical stress in any dialect apart from Inupiaq on the Seward Peninsula in Alaska, which Kaplan (1985) describes as iambic. Studies of the remaining dialects show that intensity prominence is unsystematic, consistent with the data considered here. Finally,

SL operates independently from intonation. Among the Inuit languages, the most studied tonal system is that of West Greenlandic, where SL is not operative. In that dialect, interrogative phases are marked with boundary melodies and final syllable lengthening.

Massenet (1980) describes a similar system of intonation for Quebec Inuttitut, where SL is operative. The rule is not affected by these tonal patterns, consistent with the results discussed here in Chapter Five.

69 4 Methodology

4.1 Preliminaries

The primary source data for this thesis come from field work performed in 2009 and

2010. After gaining ethical approval from Memorial University and a research permit from the Government, I conducted 19 days of field work on the North Coast of Labrador in April of 2009. This included radio interviews with the 0 KalaKatiget

Society to publicize the research, canvassing of potential Labrador Inuttut speaking language consultants, ethnographic interviews and continuing the preparation of my linguistic research program. Field work resumed in April of 2010 for 21 days of linguistic interviews with language consultants and travel on the land with cultural experts in and around the communities ofNain, Hopedale, Makkovik and Rigolet. In each town, local authorities were consulted and information sessions were held at community halls

(attendance: Nain 17; Hopedale 15; and Rigolet 12). Information sessions were also held at three local schools, involving more than 60 students.

4.2 Participants

More than thirty people were directly involved in this study, including translators and language consultants, in addition to their spouses and other family members. For the linguistic and ethnographic interviews, examples from 22 language consultants were transcribed and segmented into a corpus organized by the software program Phon. From this corpus, examples were then measured with the speech analysis software program

70 Praat. Nine language consultants were women (AE, BH, DF, FW, JD, KT, MH, VI and

SI) and 13 were men (AZ, BK, EF, HP, HW, JI, JM, Ll, MK, MN, PA, PJ and TK).

Language consultants ranged in age from 41 to 81 years old. Based on their answers to questions about their parents and grandparents, the language consultants were categorized as being from speech communities roughly divided as follows: Inuttut ((Hebron 9), (Okak

3), (Nain 6), (Hopedale 3)) and Inuktut (Rigolet sub-dialect 1).

4.3 Audio environment and equipment

Transportation was an issue for some of language consultants, so quiet areas were established in their homes with appropriate mixing and microphone placement to capture high-quality recordings. Alternately, interviews were done in a quiet area set up at the

Atsanik Lodge, a hotel in Nain. Two interviews were done at the OKalaKatiget Society's

Broadcast Centre in Nain. Language consultants from Makkovik were interviewed in their homes. In Hopedale and Rigolet, language consultants were interviewed either in their homes or in a quiet room set up in two local hotels. All interviews were documented with the following professional equipment: Sony DV camcorder, M-Audio Microtrack

24/96 audio flash recorder and an HHL professional MD recorder. Main source audio came from a boomed Electro-voice RE-50 microphone.

4.4 Goals and methodology

The primary goal of the field work was to gather examples of SL in the speech of

Labrador Inuttut speaking language consultants living in Nunatsiavut, the Inuit-governed

71 part of Labrador. My hypothesis was that the durational pattern of SL as described in the literature would emerge from such data. Secondarily, I wanted to investigate the possibility of an iambic metrical pattern, or remnant thereof, since one of the key examples of iambic footing in MST comes from the related but more conservative

Eskimo-Aleut languages of the Yupik (Hayes 1995). The ultimate goal was to compare the metrical system in Labrador Inuttut data with the occurrence of SL, with the working hypothesis that they would operate independently, as suggested by Dresher & Johns

( 1995).

4.4.1 Design and implementation of the linguistic interview

With help from my academic supervisors, I designed a linguistic interview to elicit the required data. A trial of that linguistic interview was done at Memorial's

Speech Sciences and Language Acquisition Laboratory with a language consultant who preferred to remain anonymous. Analysis of that data offered the basis for the final revisions of my linguistic interview materials, provided in Appendix A. Each interview began with a conversational exchange intended to establish background personal information. The results for Section A of the interview process provided a series of words used by all language consultants. An important example word used in this study is

[ana:naya] 'my mother' which arises in multiple conjugations and phrasal positions

(described in §4.4.4), making it an appropriate token for acoustic testing on the possible correlates of stress, presented in §5.3.

72 Sections B and C of the interview involved reading tasks. Language consultants were first asked to read from the top of the seventh page of Labradorimi Ulinnaisigutet, which is the introduction of a dictionary published by the Labrador Inuit Association in

2006. Next, language consultants read from a series of words, minimal pairs and lexical items with SL alternations, each embedded in carrier sentences. Since the durational

1 pattern of SL is represented in the Labrador Standardized Spelling System (LSSS) , these reading tasks did not constitute a neutral test for the application of SL. None of the results for Sections B and C were thus considered in this thesis in the context of SL.

Section D is the / tutuk/ 'messy hair' versus / tuttuk/ 'caribou' oral task. Language consultants were asked to make up four sentences, each with a different example word based on either [tutuk] or [tuttuk]. These words were given to the language consultants orally. They were also not allowed to see the written form given to the translator. Instead the translator was instructed to explain the meaning of the word without actually using the construction. In the case of bilingual language consultants, the meaning of the word was explained entirely in English. For example, with [tuttu:IJuaqxauiJittuk], language consultants were asked to imagine a situation in which they would use a word that means:

"she wasn't pretending to be a caribou earlier today." Despite the somewhat artificial nature of the task, without exception language consultants produced all four alternations without the aid of the written form shown in Section D. The results from this task

Adopted in 1980 by a conference of Elders held in Nain. LSSS is almost phonemic, except the long vowels [a: i: u:] are written as 'a', 'e' and ' o', respectively, while the consonants [X] and [y] are written with a small upper-case letter ' K' and the lower-case letter 'g', respectively.

73 (described in §4.4.1) are crucial to the demonstration of the exceptionless nature of SL, in

§5.1.

In Section E, consultants were asked topical questions about life in Northern

Labrador. The results varied, with some language consultants giving long and detailed answers while others gave only short responses. Within these results, several words were repeated, but none specifically appropriate to a study of SL and syllable prominence. Of interest from this data set are the parts where the language consultants became emotionally engaged in what they are describing, something that occurs in particular for questions one and four. The difficulty is that the lexical items used in these responses are too varied to form tables of intra-speaker examples. As a result, there only a handful of example words from Section E in this thesis, 2 however these results do provide examples of utterances and speech samples important to the discussion of Labrador Inuttut prosody in §5.4.

4.4.2 Ethnographic interviews: purpose and methodology

I first took an interest in Labrador Inuttut in part because of the surprising realization that, even though the dialect has been ' documented' in terms of having a proper dictionary and grammar, many forms describing the natural environment, travel and hunting have not been written down. The existence of these forms is threatened because Labrador Inuttut is the mother tongue of so few Inuit under 40 years old. It is one outcome of colonization,

2 Numerous examples of the token, / xattax/ 'repeatedly' from EF were extracted for this thesis, with the results discussed in §5.2.

74 first by German-speaking missionaries in the 18th century to Labrador, then by English­ speaking immigrants. In the 1950's the provincial government ofNewfoundland started to enforce an education strategy that saw the relocation of Labrador Inuttut speaking children away from their parents for long periods to English-only boarding schools.

Nunatsiavut, the Inuit government in Northern Labrador, is now trying to revitalize their language. Fallowing their example of holding story telling gatherings and land-based language camps, I endeavoured to include elements of the same in my linguistic study.

The first field trip to Labrador in 2009 was in part reconnaissance for the linguistic interviews and an opportunity to do ethnographic research which in the end provided data crucial to this study. Background work was first done for two ethnographic studies: first I compiled a list from the Labrador Inuttut dictionaries of all the words related to sea ice; and I then compiled a second list of all the species of flora and fauna named in the biology literature. The latter was correlated to a folder of photographs printed from

Google Images for each species. To obtain photographs of sea ice conditions, trips were taken on snowmobile over the frozen sea around the communities. From Nain, a trip was undertaken with three Inuit guides to an [iiJiyyanik] 'polynya' 30 kilometres away. From

Rigolet there was a trip with two Inuit guides to two open water polynyas: one on a river and a second on the sea. Finally, from Makkovik, an Inuk father and son took me by snowmobile to the [sina:] ' sea ice edge', 15 kilometres off shore. Hundreds of sea ice features were photographed and correlated to words in the Labrador Inuttut lexicon. One feature, for example, an [allu] 'seal's breathing hole', was found near Makkovik. The resulting discussion in Labrador Inuttut about the signs left in the snow and ice by a seal

75 was recorded. During the ethnographic interviews, these videos and photographs3 were presented systematically to each language consultant. The resulting spontaneous discussions included explanations, descriptions and stories about the sea ice, plants and animals. The results are important to this phonological study because they contain intra- speaker sequences of /-ixxi-/ from the base-stem, /axixxik/ 'ptarmigan' .4 These 52 tokens are directly relevant to this study, as each contains an intervocalic geminate. My hypothesis was that if the morphology supplies an underlying geminate in the next adjacent sequence to /-ixxi-/, SL must apply. Where SL applies vacuously, any prominence pattern related to syllable weight should be visible, the subject of §5.3, in the following chapter.

4.5 Software and settings

Measurement of the relevant sequences was done with the phonetic analysis software suite Praat, version 5.1.19. For the analysis of pitch, the following settings were used: time step 0.1 , pitch floor of 75 Hz to pitch ceiling 300Hz for male language consultants, pitch floor of 100 Hz to pitch ceiling 500 Hz for female language consultants. Formant settings were as follows: Maximum Formant Hz 5500, Number of Formants 5, Dynamic range dB 30. Intensity settings, View Range dB 50-100, Mean Energy averaging method.

Finally, the spectrogram setting was a dynamic range of dB 75.

3 As well as photographs of sea ice conditions in the Bering Strait (Krupnik & Weyapuk 20 I 0). 4 Produced variably by speakers as /-iyyi-/, /-ixxi-/ or / -ikki-/ the result ofthe optional phonetic implementation rules: devoicing or continuancy.

76 4.5.1 Acoustic analysis

As mentioned above, one of the primary goals of this field work was to find some pattern of syllable prominence, possibly related to the iambic pattern shown for Yupik. Word examples were thus extracted from / tutuk/ versus / tuttuk/ and three sequences in the ethnographic interviews from the base stems /ana:na-/ and /ax.ixxi-/ and the suffix

1-x.attax.-1. The tokens were first transcribed into the Phon software program, designed at

Memorial University to facilitate the phonological analysis of data. Each example was then exported to Praat for measurement. Using both auditory and visual cues from the spectrogram, the three correlates of stress were considered: duration, fundamental frequency and peak syllable intensity (Fry 1958, Liberman & Prince 1977, Hayes 1995, inter alia). Values for each were then entered back into Phon for each segment under tiers for loudness, duration and pitch. This was then exported into the tables which appear in

Appendices C-F. A graphic representation of the three stress correlates for some of the examples was created with Praat as a visual aid in the description of results in Chapter

Five.

4.5.2 The phonemic pair /tutuk/ 'messy hair' versus / tuttuk/ 'caribou'

The data set in Appendix C was designed to test for SL in spontaneous speech. It resulted in 32 usable examples, some discussed already in the preliminary results (see §2.2.1 ).

There were eight language consultants involved of various ages: five women (AE-69, JD-

53, BH-56, MH-41 and FW-63) and three men (PA-46, MK- 43 and TK-51). Most of

77 these example words arise in phrase-medial positions, but a few occur phrase-finally.

There is systematic lengthening of the final-syllable rhyme, a phenomenon that will be explored in §5.4 using a different data set. For the analysis of ltutukl versus l tuttukl in

§ 5.1, which focuses on the SL durational phenomenon, this lengthening phenomenon is left aside: the duration of final-syllable coda consonants is not included in the results.

This approach received support from the consideration of word edge effects in the literature on West Greenlandic. Nagana-Madsen (1992: 118) observes that the It/ in the ultimate syllable of l ata:tal and lata:ta:l is longer than It/ in the penultimate syllable of lata:ta:tal and lata:tattal. She atrributes this length to a pattern of 'prepausal lengthening'; in §5.4, using data from the ethnographic interviews, I will show that a similar lengthening phenomenon covaries in the data considered here with pitch effects, specifically boundary melodies.

4.5.3 The morpheme lxattaxl 'often, intermittently'

The data set in Appendix D is also discussed in Chapter Five. It involves a single language consultant (EF, 69 years old, male), combining 18 example words from both the linguistic and ethnographic interviews. This data set was compiled with the hypothesis that a prominence pattern might emerge from the way one speaker uses the morpheme l xattaxf 'often, intermittently', a derivational affix often found in the spontaneous data considered here. The sequence is crucial to this study because, unlike the l tutuk/ versus l tuttukl example, the intervocalic geminate arises between the short vowel [a] , left-

78 adjacent to both the SL trigger and target. Intensity for this data set patterns unsystematically, providing additional evidence in §5.2 for the hypothesis that SL is not governed by syllable prominence.

4.5.4 The sequence [xixxi]

The data set in Appendix E includes 52 example words extracted from the ethnographic interviews with 15 language consultants of various ages: five women (AE-69, BH-56, SI-

79, VI-74 and KT-53) and ten men (EF-69, JI-73, LI-70, PJ-49, BK-53, JM-63, MN-61 ,

HP-81 , HW-72 and AZ-46). The sequence occurs within five lexical items describing related bird species, generally called 'partridge' in Labrador English. Part of the exercise was to try to better understand the Labrador Inuttut names for all flora and fauna, a subject of some confusion in the published dictionaries. Each language consultant was thus shown a photograph of the species, and, without using any assumed names, invited to make comments. Responses varied from a simple statement of the species name as they knew it, to long and detailed descriptions. As a result, word examples arise in isolation, in phrase-medial positions or in phrase-final positions. The impact this has on the observed values for all three stress correlates of the intervocalic geminate, adjacent vowels and overall word examples is the subject of §5.3.

79 4.5.5 The sequence [na:na]

The final data set, in Appendix F, is based on nine word examples extracted from a question in the linguistic interview where the responses included spontaneous phrases with the example word [ana:naya] 'my mother'. The resulting data table includes nine example words from seven language consultants of various ages: three women (JD-53,

DF-61 and BH-56) and four men (PA-46, MK- 43 , TK-51 and MN-61). The isometric nature of the CV:CV sequence makes it an ideal test for any recurring syllable prominence pattern. As with all the data discussed in Chapter Five, these example words pattern unsystematically for all three stress correlates.

80 5 Project data: results and analysis

5.1 SL: acoustic manifestation

In this chapter I present the results of my 2009-2010 fieldwork in four Labrador Inuit communities. As we will see, SL is exceptionless in the spontaneous speech of my consultants, consistent with other descriptions in the literature. This generalization is supported by the measurement of consonant duration in 32 example words where the base-stems / tutuk/ and / tuttuk/ are morphologically concatenated with which, together, yield underlying sequences of geminates adjacent to one another across a single syllable nucleus. Nowhere in these data do syllable-adjacent geminates arise at the surface level. Instead, phonetically realized geminates can be adjoined maximally by a short consonant. These results also show inter-speaker variation in constructions involving the morphemes / -IJI)it-/ 'negative' and / -tuk/ '3s'. Despite these unexpected results, none of the variations violate SL.

5.1.1 Fieldwork data consistent with previous descriptions of SL

In this section I discuss findings from the / tutuk/ and / tuttuk/ alternation task, described in §4.5.2 as a way to test for the prevalence of SL in spontaneous speech. Recall from

Smith's (1975:105) aural-impressionistic description, SLoccurs in all cases: "there are no sequences of the form ...CC(V)VCC ... ". This generalization holds for all32 example words with the base-stem / tutuk/ or / tuttuk/ . Before discussing SL's acoustic manifestation in the data, consider first the criteria used in the representation of segments

81 as geminate. Since the / tutuk/ and / tuttuk/ alternation task involves the phonemic pair

/ tutuk/ 'messy hair' and / tuttuk/ 'caribou', one way to investigate geminate duration is the systematic comparison of the [t] and [tt] segments. Consider the following four examples (see Appendix C for all of the relevant data): 1

(1) a. [tutu:IJIJuaxauiJIJituk] messy.hair-be-pretend-NEARPAST -NEG-3s 's/he did not pretend to be messy hair' t u t u: I) I) 'ua X au I) I) t u k 28 61 67 111 108 124 91 98 121 78 81 95 76ms BH

b. [tuttu:IJuaxaul)ituk] caribou-be-pretend-NEARPAST-NEG -3s 's/he did not pretend to be a caribou' t u tt u I) 'ua X au I) i: t u kh 19 88 205 87 78 227 63 127 88 129 21 91 140 BH

c. [ tu tu:IJIJ uaniaiJIJituk] messy. hair-be-pretend -NEARFUTURE-NEG-3 s 's/he will not pretend to be messy hair' t u t u: I) I) 'ua n ia I) I) i: t u kh 29 75 52 120 114 101 63 126 111 99 58 97 125 BH

d. [tuttu:IJuaniaiJIJituk] caribou-be-pretend-NEARFUTURE-NEG -3S 's/he will not pretend to be a caribou'

t u tt u: I) ua n 'ia I) I) t u k 52 67 180 146 59 128 73 128 124 99 56 43 103 BH

The use of apostrophes in this table and all those that follow, as in (I a) ['ua] , denotes the syllable with peak word intensity. The initials at the far right denote the language consultant.

82 In (1a), the singleton [t] has a duration of 67 milliseconds (ms hereafter) while in (1 b) the geminate [tt] is 205ms. The geminate is thus 3.1 times longer than its singleton counterpart. In (lc, d) the geminate is 3.5 times longer. On average [tt] (198ms; n=27) is

3 times longer than [t] (66ms; n=66) for the examples in Appendix C. This is consistent with all the data considered here as well as with comparable data from acoustic studies of

West Greenlandic. Mase & Rischel (1971 :235) and Nagano-Madsen (1992:61) argue that long consonants are "two times" longer than short consonants. Jacobsen (2000) also finds a systematic durational difference; she argues that a long segment (where 'segment' stands for either a consonant or a vowel) is shorter when preceded by another long segment (as described in §3.5.2). Jacobsen (200:60) compares [kk] (248ms; n=2) in

[kuk.ukk.umavara] with the singleton [k] (136ms; n=2) in [kuk.uk.ulava:t]. On average then for Jacobsen's two West Greenlandic language consultants, the geminate is 1.8 times longer than its singleton counterpart, consistent with the data considered here. These examples support the representation of geminate consonants in the following discussion ofSL.

Recall that the objective set in §4.5.2 for the / tutuk/ and / tuttuk/ alternation task was to observe the behaviour of morphemes in opposite environments: following a syllable with an SL trigger versus following a syllable where SL is not a factor. Smith

(1977a, 1978) describes the morphemes selected for the current investigation as / -IJIJUa-/

'to pretend, play x', 1-xxau-/ 'near past' and /-IJIJituk/ 'negative.3s'. According to Smith, these morphemes alternate under the influence of SL with the degeminated surface forms,

83 [IJua], [xau] and [IJittuk]. Beginning with the morpheme / -IJIJUa-/, the degeminated form arises in all cases for the base-stem / tuttuk/ and never for / tutuk/, as shown in the following examples, with segments represented as geminate (SL triggers) shaded in light grey and segments represented as degeminated (SL targets) shaded in dark grey. The durational ratio between the trigger and target, here called the geminate shortening ratio

(GSR), is shown for each instance of SL in the following examples:

(2) a. [tutu:IJIJUaxaul)ittuk] messy.hair-be-pretend-NEA RPAST-N EG-3S t u t u: IJIJ 1Ua au IJ tt u th 44 128 61 160 150 166 178 57 82 117 60 263ms FW GSR: 1.2

b. [tuttu:uuaqxauuituk] caribou-be-pretend-NEARPAST-N EG -3S t u tt lu: ua QX au t u kh 64 50 258 190 158 210 174 103 88 34 380 FW GSR: 2.7 GSR: 3.9

c. [tutu:IJIJUaniaiJittuk] messy.hair-be-pretend-NEARFUTURE-NEG-3s t u t Ill IJIJ ua n ia IJ tt u: xh 68 97 39 13 164 182 90 111 86 70 120 250 553 MK

d. [tuttu:uuaniaiJIJituk] caribou-be-pretend-NEARFUTURE-NEG-3S t u tt lu: ua n ia IJIJ u k 57 45 255 161 176 67 140 117 60 48 96 FW GSR: 2.9 GSR: 1.4

84 In (2b, d), [IJua] arises following the SL trigger [tt]. As we can see in (1) this pattern is, again here, exceptionless. Observe further that, on average in Appendix C, [IJIJ] ( 145ms; n=24) is 1.6 times longer than [IJ] (89ms; n=39) consistent with the representative results in

(2). Crucially, in spite of considerable inter-speaker and intra-speaker variation in the duration of phonemically long versus phonemically short segments, their respective values never overlap.2 This is consistent with Jacobsen's (2000:42) account of West

Greenlandic. In the remainder of this thesis I, will refer to the above calculation as the

'geminate shortening ratio' (GSR), which expresses how many times longer the SL trigger is, when compared to its target.

With the /-xxau-/ morpheme, the degeminated form [xau] arises in (2a) where it follows the SL trigger [IJIJ]. By contrast, in (2b) , where SL is not a factor, the surface form is [qxau]. This pattern holds for all the examples (listed in Appendix C), which provides further evidence that SL is exceptionless. Also, the fact that SL applies twice in the examples in (2b, d) is consistent with the claim that SL is an iterative rule. In (2b) the rule applies in consecutive adjacent syllables. In (2d), however, the pattern fails to apply as the relevant context contains a morpheme that lacks an SL trigger, the underlying / -nia(x)-/

'near future' (Smith 1978:77). That allows the [IJIJituk] form to arise in the adjoining

2 At its lowest, the geminate shortening ratio is 1.2, as in (2a) (see also Appendix C, ( I b). If we consider speaking speed, notice that one of these two examples, (2a), is the second fastest sequence for the paradigm / tutu:IJIJuaxaul)l)itu/ at 1.07s, while the other comes from a slower speaking speed example, (I b) in Appendix C at 1.33s. It thus cannot be said that lower geminate shortening ratios associate exclusively with faster speech rates, though this is more often than not the case in the 32 Appendix C examples. All of the instances ofGSR 1.3 in Appendix C occur in the fastest, second and third fastest speech rate examples for their paradigm: (I a) at 0.97s, (2b) at 1.05s, and (3c) at 1.1 2s.

85 position. 3 As a result, the underlying form of (2d), prior to application of SL, is / IJIJittuk/ .

Observe that the iterative pattern of SL resumes, despite the intervening non-SL trigger syllable. This finding is consistent with the position in the literature that SL is an exclusively syllable adjacent phenomenon. Finally, in (2c) the /-IJIJit-/ and / -tuk/ morphemes do not follow the pattern in (2d), the case for a number of examples in

Appendix C. This morphological variation will be discussed further in the next section.

For our purposes here, observe that none of the variations involving these morphemes result in an SL violation, the unattested surface form * [IJIJittuk].

In sum, SL truly is exceptionless: there are no SL violations in these 32 examples of semi-directed spontaneous speech.

5.1.2 SL holds despite variations of the 3s negative morpheme

Smith (1978) describes the morpheme / -IJIJit-/ as a deleting suffix (see §3.2.3), in other words causing the deletion of a base-stem final consonant if there is one. He describes

/-tuk/ is an adjoining affix, in other words adjoining a base-stem final consonant if there is one. In a footnote he describes widespread inter-speaker variability in the application of deleting versus adjoining rules. The language consultants interviewed for this thesis were asked about the [IJIJituk] / [IJittuk] alternation and several described confusion about the proper usage. The results also show intra-speaker variation. Consider the following:

3 This alternation is more complex than Smith's description, as it involves two morphemes: / ·IJIJit-/ ' negative' and / -tuk/ '3s' (Douglas Wharram, p.c., October 2010).

86 (3) a. / tutu(k) + U + IJIJUa + XXaU + IJIJi(t) + tuk/ messy.hair-be-pretend-NEARPAST-N EG-3s t u t u IJIJ ua ' au IJ t u k 62 69 80 22 200 172 93 124 45 128 23 108ms MK

b. / tuttu(k) + U + IJIJUa + XXaU + IJIJi(t) + tuk/ caribou-be-pretend-NEARPAST-NEG-3s t u tt u ua X ' au IJIJ u t 36 93 139 49 237 103 85 143 44 31 60 MK

c. /tutu(k) + u + IJIJUa + nia(x) + IJIJi(t) + tuk/ messy.hair-be-pretend -NEARFUTURE-NEG-3 s t u t u IJIJ ua n 1a IJ tt u: xh 68 97 39 13 164 182 90 Ill 86 70 120 250 553 MK

d. / tuttu(k) + u + IJIJUa + nia(x) + IJIJi(t) + tuk/ caribou-be-pretend-NEARFUTURE-NEG-3S t u tt u ua n 'ia IJ t u kh 96 39 184 68 191 70 131 Ill 62 129 29 818 MK

The examples in (3a, d) show a surface form not described in the literature: [IJituk]. In

(3b) the expected form arises, while [IJittuk] is the surface form in (3c). As with (2c) above, it is not clear why in (3c) the underlying morpheme / -IJIJit-/ is degeminated in the surface form. It may be that diachronic changes addressed by Smith (1978: 116) are ongoing. As for the forms in (3a, d), these are two of the fastest performances of this particular word form, which suggests some 'flattening' of geminates in faster speech.

These speculations however lie beyond the scope of my study. Lacking the data to fully understand the cause of the variation within the phonetics of this context, the important

87 point for the current study is that none of the unexpected variations of this affix involves a violation of SL.

5.1.3 SL operates independently from vowel length

In this section, I show that, consistent with the literature, the length of the vowels adjacent to the consonants modified by SL is irrelevant to the operation of the rule. From the 32 example words in Appendix C, we can see 38 instances of degemination. In 27 of those cases, the segment in the intervening rhyme is either a long vowel [u:, i:] or a vowel sequence [ua, au]. In the other 11 cases, the segment in the intervening rhyme is a short vowel [u, i]. There is thus no evidence from these examples that vowel length makes a difference to SL, consistent with Dresher & John's (1995:81) observation that long vowels are "no impediment to the operation of SL." As discussed in §3.3.2, Dresher &

Johns (1995 :89) take this argument further, concluding that "just as vowels do not affect

SL, they are never affected by it either. Thus there is no vowel shortening in the context of SL." Evidence in support of this position comes from the following examples.

Consider the second syllable in each which, following Smith (1978), is underlyingly

/ tuttu + u/ ' be a caribou' or / tutu+ u/ ' be messy hair' :

88 (4) a. / tuttu(k) + U + I)I)Ua + XXaU + I)I)i(t) + tuk/ caribou-be-pretend-NEARPAST-NEG-3S t u tt u: 1Ua XX au tt u k 21 38 125 166 243 119 117 70 125 27 56 MH

b. / tuttu(k) + U + I)I)Ua + XXaU + I)I)i(t) + tuk/ caribou-be-pretend-NEARPAST-NEG-3S t u tt u 1Ua qx au tt u kh 47 101 140 101 180 120 98 74 191 100 153 PA

c. / tutu(k) + U + I)I)Ua + XXaU + I)I)i(t) + tuk/ messy.hair-be-pretend-NEA RPAST -NEG-3 s t u t u I) I) ua 1au I) t u k 62 69 80 22 200 172 93 124 45 128 23 108 MK

d. / tutu(k) + U + I)I)Ua + XXaU + I)I)i(t) + tuk/ messy.hair-be-pretend-NEARPAST-NEG-3s t u t u: I) I) 1Ua au I) t u kh 26 37 126 120 180 186 124 127 34 121 28 302 MH

Observe that in ( 4b, c) the second syllable peak is a short vowel. This is the case for 10 of the 32 examples in Appendix C, a morphological variation that may be attributable to the spontaneous nature of the oral task. For some reason, a few language consultants dropped the -u- 'to be' morpheme.4 The motivation for this will not be explored here. Again, the important point is that the different vowel lengths of the second syllable in ( 4a, b) have no impact on the application of SL, since the underlying geminate / IJIJ/ is degeminated in both cases. Note as well that SL has no systematic impact on the length of a syllable peak it straddles; as discussed above, the difference in vowel length in ( 4a, b) is morphological.

4 The results are still grammatical words, though somewhat artificial in meaning.

89 In (4c, d), observe that a short vowel versus a long vowel in the syllable peak preceding a

SL trigger has no impact on the degemination of an SL target, in this case l XX/. These general patterns described for (4) hold without exception for all the data considered here.

5.1.4 Summary of durational results for /tutuk/ versus /tuttuk/

The encompassing conclusion is thus that SL is exceptionless in environments where the morphology supplies syllable-adjacent underlying geminates. Generally from the data in

(1-4) and the 32 examples in Appendix C, phonemically long consonants, [tt] (198ms; n=27) and [IJIJ] (145ms; n=24), are 2.2 times longer than their phonemically short counterparts, [t] (66ms; n=66) and [IJ] (89ms; n=39). In the context of SL, geminates are at the very minimum 1.2 times longer than the underlying geminates they reduce. On average for the data in Appendix C, sequences that trigger the rule, [IJIJ] ( 151 ms; n=20),

[tt] (209ms; n= 16) and [qx] ( 167ms; n=6) are 2.1 times longer than the underlying geminates they target, [IJ] (70ms; n=26), [t] (77ms; n=8) and [X] ( 102ms; n=8). Some morphological variation was observed for the data in Appendix C involving 'negative.3s' and 'to be', but these never yielded SL violations. It can therefore be said of SL that morphology is only a relevant factor insofar as it provides syllable-adjacent underlying geminates. Finally, the presence or absence of SL has no systematic impact on vowel duration and a long vowel cannot block an SL trigger. The acoustic data in Appendix C thus fully demonstrates the productivity of the phonological rule, which also fully supports the descriptions found in the literature. As a dissimilation process, SL is

90 typologically distinctive in that it is blind to syllable length and targets underlying geminates. For example the Lex Mamilla rule in Latin, discussed in §2.3 .1, is like SL in the targeting geminates, except the rule operates regressively. But Lex Mamilla also refers to vowel length and syllable weight. We have seen that vowel length is irrelevant to SL.

In §5.2 syllable weight is shown to have no effect on SL either. So SL is crucially different from a rule like Lex Mamilla. Another language that might provide an analogous length dissimilation process comes from two studies of loanwords in Japanese (I wai

1989, Wade 1996). They describe the diachronic rule as follows: loanwords with lax vowels are followed by geminate plosives (e.g. 'zipper' becomes [jippa], ' lucky' [rakki:],

'platform' [purattohoomu]). The rule is not followed in cases where the loan word has a lax vowel and a geminate, as the following data from Ito & Mester (1998:23) show:

(5) Exceptions to the consonant length rule in Japanese loanwords Ciketto *Cikketto 'ticket' kecappu *keccappu 'ketchup' mapetto *mappetto 'Muppet' ootomaCikku *ootomaccikku 'automatic' maikurosukopikku *maikurosukoppikku 'microscope'

Consider the unattested forms in the middle column: the banned environment involves syllable adjacent geminates, the same environment banned by SL. The examples in (5) also show that the exception applies to geminate consonants for all places of articulation, 5 the same as the Labrador Inuttut data studied here. The only difference from SL is that, as with Lex Mamilla, the rule applies regressively. This study therefore contributes to

5 Ito & Mester ( 1998:23) show the exception *kk. .. kk also applies but their example, [piknikku] ' picnic', the hypothetical form [*pikkunikku] appears to breach syllable adjacency. Further investigation of this apparent counter-example is beyond the scope of this thesis.

91 previous typological surveys of dissimilation (Suzuki 1998 and Bye 2011) by showing another type of language with length dissimilation. In the case of the Inuktitut dialects with SL, the rule can be characterized as follows: /CC. ..CC / ~ [CC ... C] when the trigger and target are syllable adjacent within the domain of the phonological word. Crucially this last condition makes Labrador Inuttut unlike the other length dissimilation examples in Suzuki(1998: 157) since the domain of dissimilation in Finnish, Gidabal, Japanese,

Latin (Lex Mamilla), Oromo and Slovak (Rhythmic Law) is the phonological foot. As will be shown in my discussion of prominence in §5.3, there is no acoustic evidence in the corpus of data compiled for the current study suggesting a foot structure is operative in Labrador Inuttut.

5.2 SL: No metrical motivation

The durational results discussed thus far are unsurprising given the description of SL in the literature. Another key point comes from Dresher & Johns ( 1995). They convincingly argue that SL is unrelated to any system of metrical stress, though admitting that the precise nature of the rhythmic system of the language is, at best, only partially defined

(see §3.3.2). Previous research discussed in §3 .5 shows that fundamental frequency, which plays a role in the intonational system of the language, has no metrical foundation.

Similarly, recent research discussed in §3.4 shows that intensity prominence has no metrical basis in West Greenlandic, consistent with my preliminary results §2.2.2. In this section, I show evidence that, consistent with Dresher & John's (1995) observations (and

92 related analysis), SL does not co-vary with a system of metrical stress based on intensity or pitch prominence.

5.2.1 SL: unrelated to syllable intensity

In addition to the references above, evidence that intensity is unsystematic (§2.2.2) and unrelated to SL comes from ethnographic and linguistic interviews with one language consultant (see §4.5.3). The following results for duration, representative of 18 word examples in Appendix D, first show that SL is active and exceptionless:

93 (6) a. / kattu(k) + xxi + vallia + tuinna + xatta + tu(k)/ come.together-instance.of-increasing-only-repeatedly-3s k a v a 11 nn a a tt tt uI i ia I ui X al u 142 99 174 70 84 60 51 179 138 101 131 71 37 33 172 64 6lms

b. / xaiiJIJU(k) + u + ni(X) + ya + ttau + tu(k)/ polynya-be-unnoticed-HABITUAL-PASS-3S X ai I)I) u: n y a tt au X a tt a u:h 132 122 187 127 63 45 52 123 211 148 20 22 209 78 206

c. / xua(k) + yunna + niuma + xatta + tut/ freeze-be.able. to-be.expected. to-repeatedly-3p x ua y u nn a IJ iu m a x a tt a 186 204 41 59 180 68 93 140 93 47 53 38 210 73 43 147

d. / pi + xatta + tut/ do-repeatedly-3p pix a tt alu t 35 60 22 65 223 67 88 39

Limiting the analysis to the [xattatu] sequence at right edge of each word example, note first that SL occurs, unsurprisingly, in every case. The underlying morphemes are

/ xattax/ 'often, intermittently' and / tuk/ '3s' or / tut/ '3p' (Smith, 1978:88,108) . SL thus applies to the second geminate / tt/ in / xattax + tuk/---/xattaxtuk/---/ xattattuk/ (through

Regressive Assimilation) --- [xattatuk] (through SL).

The general pattern is thus a weakening of the second consonant in the context of two adjacent eve syllables. If SL and intensity prominence were to co-vary, one would expect that the first eve syllable in the [xattatu] sequence would attract loudness in a

94 systematic way. The fact that this does not occur is clear from the following intensity results for the word examples in (6):

(7) a. [kattuxivalliatuinnaxattatu] ' it is often ·ust more and more (sea ice) coming together' k a tt u v a 11 ia ui nn a X a tt a II u 60.9 70.7 63.8 73.4 65.5 72.3 67.8 71 .5 68.5 73.4 61.7 70.2 68.7 69.2 65.9 63 58 68.6 64.1 65.4dB

b. [xaiiJIJU:niyattauxattatu:] 'polynyas are often unnoticed' X ai IJIJ u: n y a tt au X a tt a u: 65.4 73.6 71.2 72.8 70.8 71.6 68.8 73.6 64.5 70.4 67.9 64.9 62 69.3 •59.6 66.8 c. [xuayunnaiJi umaxattatut] ' they are often expected to be able to freeze' X ua y u nn a IJ iu m a x a tt a • u th 63.8 80.8 68.9 71.5 71.8 75.1 71.7 72.4 69.5 71.5 64 63 .6 60.3 67.8 62 64.4 52. 8

d. [pixattatut] ' they do ( often, repeatedly' p x a tt u t 68.2 78.3 73.9 73.8 71.5 71.4 66.9 70.4 61.4

Peak intensity falls on the second syllable in [xattatu] sequence in three of the four examples in (7a-c). Only in (7d) is it possible to argue that the syllable which retains a coda after the application of SL, the [xat] in [xattatu], co-occurs with peak intensity in the sequence. However, the contrast between the 'strong' and ' weak' syllables is low, at less than 3dB. Further, peak word intensity in (7d) falls on the initial syllable. The best explanation for the data in (7) is that SL does not co-vary with intensity. The same can be said for all 18 words in Appendix D where peak word intensity is unsystematic and

95 merely suggested by mostly weak intensity differences between syllables; these data also

(weakly) contradict Dresher & John's (1995:89) claim "that on most accounts syllables with long vowels or vowel clusters have some degree of stress, typically they have the most stress" (see §3.3.2).6 Finally, words like (7d) show that syllables with short vowels

(and no coda) can also attract peak word intensity.

5.2.2 SL: unrelated to intonation

The intonational systems for West Greenlandic and Quebec Inuttitut are described in §3.5.

Similar patterns emerge from the data considered here. Acoustic results for the base-stem

/ axixxik/ 'partridge (generic, ptarmigan) or rock ptarmigan (Lagopus mutus)' show that syllables made prominent by intensity pattern unsystematically and that fundamental frequency exhibits boundary melodies at the right edge of phrases or utterances. This last phenomenon co-occurs with lengthening of the final syllable rhyme, as described for

Quebec Inuttitut (§see 3.5.2), and, in the data considered here, with final stop aspiration.

Consider the following results, representative of the 52 examples in Appendix E:

6 Presumably, this claim is related to the impressionistic relationship that exists between vowel duration and relative prominence.

96 (8) a. [axixxik] ' ptarmigan' 59 52 53 56 58 51 dB a X x x k, ]w PJ 99 74 79 97 97 56 38 ms

b. [axiyyi::kh] 'is it really a ptarmigan (in the photograph)?' 73 72 69 64 69 52 a X y y i:i kh ],P] u HP 187 49 96 76 76 254 292

c. [axixxi::] ' is it really a ptarmigan?' 57 57 66 60 70

a X x x i:i ]1p]u JM 91 79 75 109 109 489

d. [axixxivi::n] 'are you(l) a ptarmigan?' 66 63 69 65 70 69 63 60 a X x x v i:I n ] IP] u VI 87 74 65 81 81 89 120 512 153

e. [axixxivitsiavak] 'a big, pretty brooker (willow ptarmigan, Lagopus lagopus)' 70 64 67 65 71 66 71 71 73 67 69 57 a X X X v t s ia v a: kh ]oP]u SI 75 70 94 81 81 87 39 174 93 93 295 82 141 307

In (Sa), the word arises in the middle of a declarative sentence. Generally, in consideration of the data in Appendix E, it can be said that in phrase-medial environments, FO is slightly falling from left to right. Consider the following acoustic

97 results for (8a) and a pitch drawing of the highlighted area (the language consultant is male; accordingly, the pitch range setting was 75-300Hz; see also §4.5):

Figure 1: slightly falling tone

Observe the slightly falling tone from left to right in Figure 1. This illustrates the FO pattern observed for phrase-medial environments throughout the recorded data. By contrast, consider (8b) , where the word is located at the right edge of an interrogative

98 phrase and utterance. The male consultant was directed to look at a photograph of the animal species in question and responds by asking: "is it really a ptarmigan?"7 In (7b) the phrase-final syllable undergoes a strengthening phenomenon: the syllable rhyme is tripled in length and the coda stop is aspirated. The latter result is crucial because it shows that the SL trigger, the geminate [yy], does not apply to the final consonant [kh] or in any way limit the lengthening caused by aspiration. This is evidence that the strengthening rule, henceforth called Final-syllable Strengthening, applies after SL. Finally in (8b), and all the example words in the data considered here that arise phrase finally, observe the exceptionless covariance of Final-syllable Strengthening and FO boundary melody:

7 It is therefore analogous to Massenet's (1980) description of a boundary melody and "surallongement" in environments where the questioner wants clarification of a possible misunderstanding (see §3 .5.3).

99 Figure 2: HLH boundary melody

The movement of FO in Figure 2 is evidence of a HLH boundary melody, the same phenomenon described for Quebec Inuttitut by Massenet (1980:200) and West

Greenlandic by Rischel (1974:97), Nagano-Madsen (1993 :152) and Gussenhoven

(2000:133). In (7c) the male consultant is again asking a clarification-type question. The acoustic results below show FO movement familiar from Figure 2 and again co-varying

100 with the Final-syllable Strengthening effect of syllable rhyme lengthening (since the word in (8c) lacks a coda consonant, final stop aspiration is not a factor):

Figure 3: marginal HLH boundary melody

The FO movement suggests an HLH boundary tone, though not as clearly implemented as the one shown in Figure 2. The pattern predicted in the literature is thus not always perfectly represented in the data, although it generally holds. Consider another interrogative example, (8d). The word was extracted from an animated conversation: one

101 language consultant is imitating a bird 's call and the speaker asks playfully if he is a ptarmigan [axixxivi::n]. Smith (1977a:45) describes the suffix [vi:n] 'do/are you(l)?', so the syllable peak in this morpheme is underlyingly long. Realized in Figure 4 at 512ms, the final syllable peak in (7d) is one of the longest segments observed in the data considered here; it is one of the clearest examples of lengthening. Unlike (8b) and like

(8c) , the strengthening rule is this case is not accompanied by aspiration, because in this case the final syllable coda [n] is [+voiced]. In (8d), a female language consultant (pitch range setting is 100-500Hz, as described in §4.5), observe an HLH boundary melody:

102 Figure 4: HLH boundary melody

103 The FO movement is consistent with (8b), except with a more elongated HLH boundary melody because of lengthening. The final example, (8e) , also comes from a female language consultant. In this conversation, she is making animated comments to her husband about a photograph of the ptarmigan species in question. She expresses delight at the sight of this type of ptarmigan by stating how much she wants one. The utterance and declarative phrase ends with the word [axixxivitsiava:kh], based on the underlying

/axixxivik/ ' brooker' and the suffixes described in Smith (1978:103) as /-tsia-/ 'fine, well, good, properly x' and in Smith (1978: 113) as / -vvak/ 'big' . The resulting derivation is realized with Final-syllable Strengthening effects: at 149ms, the fmal vowel is doubled and the final coda stop is aspirated. Observe as well the following pitch pattern:

104 Figure 5: HL boundary melody

In this case the FO movement suggests an HL boundary melody pattern, shown at the centre-right of the pitch drawing in Figure 5, co-occurring with the lengthened final vowel. The L tone is not fully implemented, possibly because it is followed by a stop

105 coda. The pattern is nevertheless consistent with the description of declarative phrase intonation in West Greenlandic (Rischell974, Nagano-Madsen 1993, Gussenhoven

2000). The difference from those studies is the optional nature of boundary melodies, that lengthening is shown at the right edge of declarative phrases (described only for interrogative phrases in the literature on West Greenlandic and Quebec Inuttitut) and that stop consonants at the right edge of phonological phrases are aspirated, a phenomenon which, to my knowledge, is not discussed in the literature. Aspiration at the right edge of phonological phrases is found in many languages. Windsor & Cobler (20 13: 1) describe a

8 similar process in Blackfoot , while AnderBois (2009:2) shows that aspiration refers to word boundaries as in Yucatek Maya9 and Nespor & Vogel (1986:90-91) show that metrical prominence regulates aspiration in English. 10

8 For example: apiit 'sit down' l api:1/ ~ [api:d c/J. Windsor & Cobler's data on Blackfoot is unlike the Labrador Inuttut data here in that the phenomenon in Blackfoot also impacts phrase-final vowels: mistapoota 'go away' l mistapo:t.a/-+ [mistapo:ta! c/J 9 The environment for aspiration in Yucatek Maya is the right-edge of the phonological word. Voiceless stop codas in this position are aspirated, as the following data from AnderBois shows: [si:nik:] 'ant' (*si:nik, *si:nik') I 0 In English aspiration refers to footing, a level of metrical structure not observed in the data considered here. Obstruents arising in the most prominent syllable of the phonological foot are aspirated, as shown in the following data from Nespor & Yo gels (foot structure labelled with l:): sweet tooth - sweet [th]ooth [sweeth [toothh: hospital - *hospi[th]al [hospitalh: night owl - *nigh[th] owl [nighth: [owlh: Iverson & Salmons ( 1995:378) show that aspiration in English, and more generally the laryngeal feature [spread glottis], encompasses a ternary system of contrasts, as shown in the following grid (with metrical prominence of a syllable ranked (0, I, 2) according to how many asterisks it is accorded in the grid): 2 (word) * * * * I (foot) * * * * * * * 0 (syllable) * * * * * * * * [th]u nar [t]er rrainr enr [th]irer sar [t]irer The degree of aspiration is greatest in [th]una because the obstruent arises in the most prominent syllable within the foot and word. Iverson & Salmons ( 1995) argue for a secondary form of aspiration in the case of an an obstruent arising in a metrically weak position, like the onset It/ in [t]errain. This contrast is also shown in word-medial positions where the onset / t/ in en[th]ire is more aspirated than the onset I t! in sa [t]ire because it is in a metrically strong position. Finally, aspiration is absent outside the syllable

106 Overall, the evidence in (8) points to a rule in Labrador Inuttut that strengthens the final syllable of a phrase or utterance. The relevance of that finding to this thesis is the fact that Final-syllable Strengthening can be shown to occur independently from SL. For example, the underlying geminate / vv/ in (8e) is degeminated by SL /vv/ - / [v] despite being adjacent to strengthening in the adjacent final syllable.

5.3 No evidence of metrical stress in the data

So far in this chapter, I have shown that SL is exceptionless in the examples and does not co-vary with syllables made prominent by any of the three stress correlates. I have not yet discussed the nature of syllable prominence in Labrador Inuttut, however. In this section,

I will show that there is no recurring or systematic pattern of metrical prominence in the data considered here.

5.3.1 Intensity prominence does not depend on duration

The word examples considered thus far have included different vowels, coda consonants and/or Final-syllable Strengthening. To eliminate the possibility that these factors are somehow clouding the picture of syllable prominence, we must consider the results for a set of example words with no codas (and thus no SL) and only the low vowels, [a] and

[a:]. Nine examples of the relevant word, [ana:naya] 'my mother' were extracted from the spontaneous speech sections of the linguistic interviews as they were uttered by seven

onset, like the coda I t/ in nigh[t'] owl or in the onset of a syllable with the lowest degree of metrical prominence, like the It/ in hospi[t]al.

107 different language consultants (see §4.5.5). Underlyingly, the words come from the base­ stem / ana:nak/ 'mother' and the deleting suffix described in Smith (1977a:31) as / -ya/

' .poss'. Consistent with the durational patterns described for West Greenlandic by

Jacobsen (2000), on average the long vowel in the CV:CV sequence is 1.7 times longer than the following short vowel (see Appendix E for an exhaustive description of the relevant examples). Consider the following representative sample of results for duration and peak syllable intensity (with the relevant prosody, utterance, declarative phrase or word shown in brackets: u[oP [w[] w]oP]u) :

108 (9) a. u[tuxuma w[ana:naya]wasaina al)utita:lauttuk Joshua Obed nainimiumit]u ' At the time of her death my mother, her husband was Joshua Obed from Nain.' 70 70 73 68 69 67 69 db a n a: n a y a BH 81 71 93 54 78 48 78 ms

b. u[w[ana:naya] w ma:nimiu Nainimiuk -urn- ... ata:taya: United States-imiuk]u ' My mother is from here, from Nain, urn, my father is American.' 70 72 77 72 79 75 81 a n a: n a y a JD 70 56 141 53 63 34 84

c. u(op (ata:tai ma:nimiuiJujuk ma:nimiul)uqxauju:kh]oP or[w[ana:naya:] w]oP··· op (nainmi -uh-]oP or[nainimi: tauni nainimi -uh-]oP IP [xanu:l)] IP]u ' Father is from here, was from here ... my mother. ... in Nain, uh, in Nain down in Nain. How?' 63 65 69 67 69 70 75 a n a: n a y a" PA 35 77 138 59 68 27 259

d. u(op (ata:tayalautaya:]oP op (nuta:miuyulauttu:]opoP [w[ana:naya:] wJoP op (ku:jjuamivuk panaitiluyu:]op]u ' My father was from Nutak, my mother she's Kuujjuaq and up there (in Northern Quebec)' 70 71 71 68 69 66 61 a n a: n a y a: MK 90 63 154 82 72 61 124

The antepenultimate syllable in [ana:naya] co-varies with peak intensity in (9a, d) and is an underlying long vowel. It could be wrongly argued from these facts that peak duration and intensity co-vary. That cannot be, indeed, since in (9b, c) peak intensity co-varies

109 with the final syllable, a short vowel underlyingly. Furthermore, the degree of contrast between all the intensity peaks in (9) is rather minimal, as it never exceeds 4dB. If intensity does mark syllables as prominent, this occurs not as a systematic metrical function, but instead as a way to emphasize specific points, consistent with observations previously made by Smith (1975:104). Consider for example the acoustic results in (9b), from a female language consultant:

Figure 6: Intensity prominence is discretionary

In addition to attracting peak word intensity, the morpheme [ya] is also the loudest syllable in the utterance. If this contrast is in fact significant, it is that it functions to emphasize that the language consultant is talking about her own mother. No plausible arguments can be made for a system of recurring metrical prominence involving duration or intensity from the data in Figure 6. We can further observe that the intensity peaks in this representative example range between 70 and 81 dB with no alternating pattern of

110 prominence. Instead each syllable attracts roughly the same prominence. Based on these observations, intensity assignment must: (a) refer to the syllable (as opposed to larger constituents like the foot or the word), giving each syllable in an utterance similar peak intensity, and (b) be agnostic to vowel duration (consistent with other observations about

SL in §5.1). The pattern in Figure 6 is therefore incompatible with Hayes' (1995) definition of a stress-timed language, which should be characterized by systematic intensity alternations between strong and weak syllables based on intensity prominence.

The evidence in (9b) supports the claim by Rose, Pigott & Wharram (20 12) that Labrador

Inuttut shares characteristics with syllable-timed languages (see §5.4, and see also Rischel

1974, Nagano-Madsen 1993, and Jacobsen 2000).

5.3.2 Intensity prominence does not depend on prosodic factors

More generally, prosodic conditioning fails to offer explanations for the distribution of intensity peaks observed in (9). The example in (9a) is utterance medial, while the example in (9b) is utterance initial. There is no evidence in any of the examples considered here that position within an utterance systematically impacts the distribution of intensity peaks. In (9c, d) the words are located at the right edge of declarative phrases, where their final syllables both display Final-syllable Strengthening (as shown in §5 .2).

For example, we observe lengthening of the phrase-final syllable in [ana:naya:]. In this example, we also observe a boundary melody of a type not discussed yet. Recall the description ofHL boundary melodies phrase finally in West Greenlandic in §3 .5.2.

Rischel (1974) and Nagano-Madsen (1993) present acoustic evidence of high-to-low

111 boundary melodies at the right edge of declaratives. In light of their description, consider now the following acoustic results for the utterance in (9d), from a male language consultant:

Figure 7: Evidence of a declarative phrase

The final syllable is characterized by lengthening as well as by a sharp downward movement of FO at the right edge, consistent with the pattern seen thus far for the declarative phrase in Figure 5. A closer look at the acoustics of this example word shows the FO movement more clearly:

112 Figure 8: marginal HL boundary tone

The expected HL tonal pattern, as in Figure 5, is not implemented. This suggests that lengthening alone can signal the end of declarative phrase and suggests that the HL predicted by the literature in this environment is optional in the data considered here. The

HL pattern may not be best instantiated by Figure 8 however, given the fact that in (9d) the boundary after [ana:naya:] appears to be a pause, while the language consultant

113 recollects information about his mother. Better evidence of a HL boundary melody comes from the final word in the preceding declarative phrase about the language consultant's father, as shown in the following spectrogram:

Figure 9: HL boundary tone

In line with the description of pitch in this environment from the literature, the most plausible interpretation for the falling movement of FO at the centre of the pitch drawing in Figure 9 is that it instantiates a declarative phrase-final HL boundary. When this boundary melody co-occurs with vowel lengthening as in (8c, d) it acts as a cue to the end of a declarative phrase (and possibly stop aspiration as shown for example words at the right edge of interrogative phrases in § 5 .2). Peak intensity therefore does not systematically co-vary with Final-syllable Strengthening in both word examples, since in

(9c) peak intensity falls on the final syllable while in (9d) it is antepenultimate. Overall,

114 the data in (9) shows that intensity and pitch operate independently and that pitch effects minimally occur in syllables at the right edge of a prosodic category (word, phrase or utterance).

5.4 Evidence for syllable timing

The evidence presented thus far supports the position that the data considered here are typologically consistent with syllable-timed languages, in line with the findings of

Rischel (1974), Nagano-Madsen (1993), Jacobsen (2000) and Rose, Pigott & Wharram

(2012). Three durational patterns have been demonstrated: phonemic contrast and SL, both discussed in §5.1 , and Final-syllable Strengthening (involving vowel lengthening, aspiration of stop codas and optionally, boundary melodies), discussed in §5.2. Aside from phrase-level intonational contours, the overarching generalization about these phenomena is that they all involve the length adjustment of a syllable rhyme. Phonemic contrasts involve a lexical difference in the length of syllable peaks, as in [anak] 'faeces' versus [a:nak] 'paternal grandmother', or a lexical difference in the length of consonants as in [anak] versus [annak] 'woman' . SL deletes coda consonants in alternating syllables.

Final-syllable Strengthening lengthens the syllable peak and aspirates the syllable coda if it is an oral stop. Recall from §3.4.2 that Jacobsen (2000:64) finds syllable rhyme length adjustments in example words read in carrier sentences by two West Greenlandic language consultants. Following Rischel (1974), Nagano-Madsen (1993:66) concludes that the "syllable is the relevant articulatory unit [in West Greenlandic ]."These acoustic studies demonstrate that, like the data considered here, other Inuit dialects have

115 phonological rules that regulate syllable rhyme duration, especially when the syllables in question become extra-long because of the presence oflong segments. From this perspective, SL regulates syllable rhyme duration. Final-syllable Strengthening on the other hand regulates syllable rhyme duration at the phrase level, in that case making the rightmost syllable durationally prominent relative to the other syllable rhymes in an utterance, or aspirated, or lengthened and aspirated. In the next two sections, I will show acoustic evidence that in environments where SL and Final-syllable Strengthening are not factors, syllables fall into just two length categories, short and long.

5.4.1 Syllables in example words are similar in length

The evidence in this section comes from the spontaneous speech data discussed already as the / na:na/ sequence in §4.5.5. Consider the syllable rhyme durations in the following examples:

116 (10) a. w[a n a: n a y a]w 425 ms 34 98 81 58ms TK

b. w[a n a: n a y i] w 459 83 107 86 49 BH

c. w[a n a: n a y a]w 501 70 141 63 84 JD

d. w[a n a: n a y a]w 503 81 93 78 78 BH

e. w[a n a: n a y a]w 512 77 134 63 58 DF

f. w[a n a: n a y a]w 624 95 163 69 109 MN

g. w[a n a: n a y a:] w]op 646 90 154 72 124 MK

h. w[a n a: n a y a::] w)op 663 35 138 68 PA

I. w[a n a: n a y a::] w]op 722 98 120 97 BH

Based on the physical durations in ( 10) , three classes of syllable rhyme emerge: short, long and overlong (shown in white, light grey and dark grey respectively). Short includes

117 nine [a] syllables, with an average duration of 73 ms, nine [na] syllables, with an average duration of 75ms, and six [ya] syllables, with an average duration of 73ms. Long includes nine [na:] syllables, with an average duration of 128ms, and one [ya:] syllable at 124ms.

Overlong includes two [ya::] syllables, with an average duration of 242ms. These results are consistent with Massenet's (1980) analysis of short, long and overlong syllables in

Quebec Inuttitut (see §3.5.3), especially since the overlong syllable rhymes in (1 Oh, i) co­ vary with interrogative phrase boundaries and Final-syllable Strengthening. Consider

(lOa-f), where Final-syllable Strengthening is not a factor in the word examples. Only syllable rhymes from the short and long classes remain. The long syllable rhymes all come from the lexicon as the linguistically contrastive, long form of [a]. The remaining short syllable rhymes in (1 0) arise in different word positions (initial, medial and final) and with different segments ([a] versus [na] versus [ya]), yet each is about the same length with average durations of73, 75 and 73ms respectively. At the word level then, if phonemic contrast and Final-syllable Strengthening are not factors, each syllable rhyme is assigned short duration. The phonetic realization of short is shown in (1 0) to be between

34 and 98ms. The next section I will show evidence of short, long and overlong durational classes in phrases and utterances, as well as briefly considering the syllable rhyme duration for eve.

118 5.4.2 Syllables in phrases are similar in length

First consider the syllable rhyme durations in the following response from the linguistic interviews where the language consultant was asked where his parents are from:

( 11) u(op (ata:tayalautaya:]oPop (nuta:miuyulauttu:]op op (ana:naya:]op 'My father was from Nutak, my mother... ' op(ku:jjuamivuk panaitiluyu:]oP]u ' ... also from up there in Kuujjuak (Quebec).' MK a. oP[ata:tayalautaya:]oP a t a: t a y a au t a y a: 99 47 143 26 83 55 59 54 134 98 83 22 123

b. oP [nuta:miuyulauttu:]oP n u t a: m m y u au t u: 101 69 48 124 114 127 35 69 32 139 118 134

c. op (ana:naya:]op

a n ~ n a y ~ 90 64 154 82 72 60 124

d. oP [ku:d3uamivuk. .. k u: d3 ua m v u k 49 109 139 137 75 70 93 69 38

e. . .. panaitiluyu:]oP p a n ai t 1 u y u.. 19 47 70 108 81 47 60 46 37

The utterance must be divided into four declarative phrases based on Final-syllable

Strengthening effects observed in all cases except (lid). At the right edge of that word example, there is no syllable rhyme lengthening and the final stop is unreleased. This word must therefore form a declarative phrase with the example word in (lle). The

119 duration of syllable rhymes not affected by Final-syllable Strengthening correspond with two classes oflength. Short includes seven syllables where [a] is at the peak ([a], [ta],

[ya], and [pal) with an average duration of 7 6 ms, two syllables where [i] is at the peak

([mi], and [til) with an average duration of 59ms and four syllables where [u] is at the peak ([nu], [yu], [lu], and [vuk]) with an average duration of 73ms. Long includes five syllables where [a:] is at the peak ([ta:], [ya:], and [na:]) with an average duration of

134ms, two syllables where [u:] is at the peak ([ku:d] and [tu:]) with an average duration of 137ms11 and four syllables with a complex vowel cluster at the peak ([lau], [miu],

(jua], and [nail) with an average duration of 129ms. The other length class, overlong, occurs only in Final-syllable Strengthening environments where the final syllable [yu::] has a duration of213ms.

Overall, the data in (11) suggest that each syllable class roughly corresponds to a durational range. For example short syllable rhymes, including V, ev and eve, are all between 47ms and 107ms in (11). Long syllable rhymes meanwhile, including V:, ev: and ev:e are between 108 and 179ms. 12 Thus the durational ranges for short and long syllable rhymes in (11) do not overlap, a pattern that holds for most of the data considered here. Generally, then, it can be said then that in phrases, short syllables are realized within a similar durational range with no overlap into the durational range of long syllable rhymes.

II Assuming that the duration of the affricate in ( I Od) is divided equally between coda and onset. 12 Again on the assumption that the duration ofthe affricate in (IOd) is divided equally between coda and onset.

120 The final evidence for a system of short, long and overlong syllable rhymes comes from a spontaneous speech section of the linguistic interviews. 13 One language consultant was asked to recall her memories of a tragic hunting accident in Northern Labrador in the

1970's. Consider the durational pattern in the following response:

(12) u(op(tuttuniayiasimatiuti auiJa namukkiak] oP -uh- ' While they were trying to hunt caribou going south, uh, ' op (pittualummu:::)op -uh- op(alla: piiJasut piiJasuixu:k) op op(ajulisimajuth]op ' ... through deeply drifted snow, uh, so three, three of them, they died.' op (nuti atautsi ilanna:ya matn silli th] oP oP [aijai:::] oP 'And one my friend Martin Sillit. Oh my! ' oP[ilu:natta:] oP op (inuit] oP oP[ilu:nattik la:badr taxxanimiu attutaulauttut] oP 'All the people, all of us northerner Labradorians, we were all touched,' op(asi ujiyatta piiJasuni:kh] op op(inu:xatittinikh]oP] u 'as a group by the loss of those three.' FW a. op(tuttuniayiasimatiuti... t u tt u n ia y ia s m a ti u t 42 236 67 188 187 56 31 152 43 68

b. . .. auiJa namukkiak) op au lJ a n a m u kk ia kh 215 31 42 38 210

c. op [pittualummu:::)op p tt ua u mm u"· 45 245 225

d. op[alla: pil)asut pil)asuixu:k)op a ll a: p IJ a s u t p lJ a s u1 x u: kh 80 182 232 56 96 109 51 85 59 135 172

e. op[ajulisimajuth]oP a: J u 1 s m a j u th 153 77 67 70 84 75 359

13 See Appendix A, Section E: Storytelling.

121 f. 0 p(nuti atautsi ilanna:ya ... n u t a t au ts a nn a: y a 57 72 86 130 185 56 71 70 180 188 Ill

g. . .. matn sillith]oP m a t n s 1 ll th 69 45 68 148 77 441

h. op (aijai:::]opoP [ilu:natta:]oPo P[inuit] oP ai j at: :: u: n a tt a: 184 96 134 57 111 208 117 201 286

l. op (ilu:nattik la:badr taxxanimiu... I u: n a t k I a: b a d r t a n a: n 1 m lU 80 108 55 62 52 140 43 109 48 156 111 96 236

J. ... attutaulauttut]oP a tt u t au au tt u th 64 192 44 190 160 175 47 232

k. oP[asiujiyatta piiJasuni:kh] op a s lU J y a tt a p I) a s u n 1: kh 87 164 96 54 221 98 45 75 43 135 162

I. oP [inu:xati ttinikh]op I n u X a t tt n 1 kh 36 100 40 39 202 78 61 388

The speech sample in (12) involves 12 declarative phrases as we can see from the pattern of Final-syllable Strengthening effects. Observe for example that the first DP, shown in

(12a, b), ends with an aspirated stop. The next DP in (12c) exhibits syllable rhyme lengthening at the right edge, a pattern also shown for the first DP in (1 2h). Observe

122 Final-syllable Strengthening as well in (12d), the second DP in (12h) and (12k). The remaining DPs end with aspirated stops. Given these assumptions about the prosody in

(12), consider the duration for each class of syllable rhyme.

Short includes 20 syllables where [a] is at the peak ([mat], [ya], [na], [a], [rna], [lan], and [pal) with an average duration of78ms, 20 syllables where [i] is at the peak ([si], [ti],

[pi], [li], [i], [sil], [tik], [ni], [ji] and [tit]) with an average duration of 76ms and 11 syllables where [u] is at the peak ([tut], [tu], [iu], [muk], [lu], [sut], [ju], [nu], and [su]) with an average duration of 88ms. Overall the durational range for short in this example is between 31 and 160ms.

Long includes five syllables where [a:] is at the rhyme ([la:], [a:], [na:], [ta:] and

[xa:]) with an average duration of 172ms, two syllables where [u:] is at the peak ([lu:]) with an average duration of 121ms and 10 syllables with a complex vowel cluster at the peak ([yia], [au], [tua], [sui], [tau], [ai], [lau] and [siu]) with an average duration of

183ms. Overall the durational range for long syllable rhymes is between 111 and 236ms.

Overlong occurs exclusively environments where the Final-syllable Strengthening effect of syllable rhyme lengthening is a factor. The two overlong syllable peaks [u:::] and

[ai:::] in ( 12c, h) have an average duration of 713ms. The remaining environments not discussed so far for (12) are those cases where Final-syllable Strengthening includes only an aspirated stop: [kiakh], [xu:kh] , [juth], [lith], [nuith], [tuth] and [nikh]. In terms of duration the syllable peaks in all cases pattern either with short or long, but measurement of the syllable rhyme for each overlong example results in an average duration of 441ms.

What remains unclear is whether or not the duration of an aspirated stop should be

123 included in the measurement of syllable rhyme length, a question that will not be answered in this thesis. It is clear that aspiration plays an important prosodic role in marking phrase boundaries. As discussed in §5.2.2, this is consistent with right-edge fortition in other languages like Blackfoot.

In sum, the picture that emerges from the data in (12) is that, except in environments where Final-syllable Strengthening is a factor, syllable rhymes are realized phonetically as either short or long and that these classes have similar durational ranges, though some overlapping of these ranges was observed in ( 12), especially for the duration of closed syllables. It may be that the phonetic reality of performing a coda consonant pushes a given syllable rhyme to the limit of its durational class. The importance here in an acoustic investigation of SL is the fact that these results show Labrador Inuttut to be unlike the stress-timed languages described in MST by Hayes (1987, 1995) because syllables are not organized into alternating strong and weak patterns. Suzuki ( 1998) proposes a Rhythmic Dissimilation Hypothesis, arguing that vowel/length dissimilation is inherently rhythmic and "driven by rhythmic principles governing foot structure (Hayes

1987, 1995)." In support of his hypothesis, Suzuki observes that alternating dissimilation patterns in Gidabal parallel the patterns of alternating stress, and that:

vowel/length dissimilation cases can be viewed as equivalent to the quantitative effects driven by the principle of foot optimization found in a number of languages. Suzuki (1998:206)

124 In Gidaballong vowels are shortened after long vowels as shown in the following data from Kenstowicz & Kisseberth ( 1997:321 ), adapted from the original study by

Geytenbeek & Geytenbeek ( 1971 ):

(13) Vowel dissimilation: an alternating pattern in Gidabal a. [njule-da:IJ] [nu:n-d3IJ] 'he (emphatic)' 'too hot' [bala-ya:] [gila:-ya] ' is under' 'that (locative)' b. [djalum-ba:-daiJ-be:] 'is certainly right on the fish' / djalum-ba:-da:IJ-be:/ djalum-ba:-da:IJ-be: ~ Not applicable: preceding vowel is short 1\

djalum-ba:-da:IJ-be: ~Applies: vowel preceded by long vowel 1\

djalum-ba:-daiJ-be: ~Not applicable: preceding vowel is short 1\ [djalum-ba:-daiJ-be:]

In (13a) we can see a ban on long vowels in successive syllables, consistent with the way

SL targets underlying geminates. In (13b) we see the resulting iterative pattern, which looks similar to SL, with one crucial difference: Gidabal is a metrically conditioned language, so the pattern in (13) does not depend on the trigger and target being sy liable adjacent, like SL, the alternation depends on foot structure. From the data considered in

§5.3 there is no evidence of foot structure in Labrador Inuttut. Intensity is unsystematic and syllable rhymes are roughly equal in terms of duration: either short or long, except phrase finally, where they can be overlong. These facts are reminiscent of syllable-timed languages described in Abercrombie (1967), Ladefoged (1975), Roach (1982), Kager

(1993, 1995), Ramus, Nespor & Mehler. (1999) and Mehler, Nespor & Shukla (2011 ).

125 This current study therefore contributes some insight into the possibility that Labrador lnuttut is syllable-timed. I recognize however that the exact nature of syllable-timing remains an unresolved question in the literature and suggest that a specific study of rhythm in Labrador Inuttut is required before any final conclusions can be made.

5.5 Summary

Overall, the results show that SL is both phonetically and phonologically exceptionless in

Labrador Inuttut spontaneous speech and that on average underlying geminates, the segmental units at the centre of this phenomenon, are 2.1 times longer than the segments they degeminate. The rule is not affected by phonemic vowel length contrasts nor does it co-vary with any regular pattern involving syllable duration, intensity or pitch. The data considered here further show that MST does not describe the metrical system in Labrador

Inuttut which must, therefore, be a non-stress language - and possibly syllable-timed.

This position is consistent with the acoustic evidence accumulated by this study which shows remarkable similarities in the duration assigned to short and long syllables, while a third category, overlong, is needed to describe the phonetic realizations of the syllable preceding a prosodic boundary.

126 6 Conclusion

The current work highlights the virtually exceptionless nature of Schneider's Law, based on primary corpus data from Labrador Inuttut language consultants. Linguistic interviews were conducted with more than thirty speakers across four Inuit communities. Analysis of the resulting data shows Schneider's Law to be exceptionless in hundreds of examples.

Even the unexpected inter-speaker variations described in §5.1.2 for the '3s' and 'to be' morphemes never result in a violation of this process. It can therefore be said of

Schneider's Law that morphology is only relevant insofar as it provides syllable-adjacent underlying geminates, the structural manifestation that triggers application of the rule.

Schneider's Law is indeed of typological interest as a rule that applies exclusively to coda consonants, irrespective of the length of the vowels preceding these consonants: the operation behaves exactly the same in syllables with short or long vowels. The fact that the output of Schneider's Law affects only coda consonants led to its first description in the literature (Schneider 1966) as a syllable sequencing rule involving eve, a view disputed by Dresher & Johns (1995). This thesis contributes to the debate by showing acoustic evidence from current Labrador Inuttut speech samples where syllable weight is not only independent from Schneider's Law, but also plays no systematic role in the assignment of intensity or pitch. The assignment of duration, meanwhile, is blind to the segmental makeup ofthe syllable since V, ev and eve are short while V:, eV: and ev:e are long. This contrasts typologically with a language like Latin, where Hayes

(1995 :51) shows that long-vowelled and closed syllables count as heavy. Labrador Inuttut

127 is consistent with the other class of languages under Metrical Stress Theory, exemplified by St. Lawrence Island Yupik, where Hayes (1995:51) shows that the division between heavy and light is based on long-vowelled versus short-voweled syllables.

However, unlike St. Lawrence Island Yupik, the data considered in this thesis clearly suggest that Labrador Inuttut is not a stress-timed language. Intensity prominence does not pattern with vowel length and is predominantly non-contrastive or unsystematic. 1 At the same time, pitch is shown to function as a cue to prosodic boundaries. In terms of duration, this thesis shows that phonemically long consonants are generally more than twice as long as their phonemically short counterparts.

Simultaneously, the output of Schneider's Law causes geminate shortening, with geminates shown in the data considered here to be at very minimum 1.2 times longer than their degeminated counterparts. We can also see in this thesis that this phonological operation is independent from the other durational pattern observed in Labrador Inuttut:

Final-syllable Strengthening. This rule applies exclusively to the final syllable of a phrase or utterance. It is observed here to output HL or HLH boundary melodies which can covary with either short syllables or lengthened long or overlong syllables. 2 As well, there is evidence that, without exception for dozens of examples, consonants arising in the coda position of phrase-final syllables are aspirated, a phenomenon not previously mentioned in the literature on Labrador Inuttut. The fact that ev :e syllables are preserved in the context of Final-syllable Strengthening, even when preceded by a eve

While allowing for the possibility that intensity prominence may be used by some speakers for emphasis. 2 With overlong syllables denoting an interrogative form where the information is not known to the speaker.

128 or CV:C syllable, in other words a Schneider's Law trigger, is further evidence that this process is not sensitive to syllable weight. Emphatically, Schneider's Law effectively eliminates syllable-adjacent underlying geminates, and nothing more.

For each phonetic correlate of stress examined, prominence is either clearly irrelevant or, at best, unsystematic. Labrador Inuttut prosodic phonology differs from that described for St. Lawrence Yupik. More generally, the unsystematic nature of the correlates of syllable prominence reported in the previous chapters makes this system incompatible with any of the stress-timed languages predicted within Metrical Stress

Theory. It seems, then, that the language must be syllable-timed, implying an evolution of

Labrador Inuttut to its current free or unsystematic stress system from the bounded, rhythmic stress systems found in the more conservative Yupik languages of the Eskimo­

Aleut language family. Recent diachronic research suggests that Schneider's Law, found in just three of the Inuit languages, is in fact the remnant of a lost metrical system (see

Rose, Pigott & Wharram 201+2.).

From a formal perspective, this work provides systematic acoustic evidence in support of Schneider's Law as it is described in the literature on theoretical phonology by

Schneider (1966), Collis (1970), Rischel (1974), Smith (1975, 1977a, 1977b, 1978),

Dorais (1976, 1990b), Dorais & Lowe (1982), Fortescue (1983), Lowe (1984), Massenet

(1986), Dresher & Johns (1995, 1996) and Jacobsen (2000). More generally, this thesis offers a contribution to the understanding of non-stress-timed languages. Among other matters, the overwhelming acoustic evidence accumulated by this study points to non­ trivial similarities in the durational range of short and long syllables. In tum, these

129 findings highlight the relative gradation in the duration of each syllable type, as well as contextual patterns- for example, the overlong duration of syllables observed at phrase and utterance boundaries. Such characteristics are also consistent with the description of syllable-timed languages in Abercrombie (1967), Ladefoged (1975), Roach (1982), Kager

(1993, 1995), Ramus, Nespor & Mehler (1999) and Mehler, Nespor, Shukla & (2011).

Overall, this thesis offers a stepping stone for further investigation of the prosodic typologies of languages like Labrador Inuttut which, despite showing distinctions at the level of vowel or consonant length (lexical contrasts) and having a prohibition on the occurrence of underlying geminates in adjacent syllables, fails to show any conditioning at the metrical level.

In sum, this thesis bridges phonetic and phonological lines of investigation and offers a number of benchmarks for future investigations of the syllable and higher levels of prosodic organization across languages both within the Eskimo-Aleut language family and beyond.

130 Bibliography and references

Abercrombie, David. 1967. Elements ofgeneral phonetics. Edinburgh: Edinburgh

University Press.

Alderete, John. 1997. Dissimilation as local conjunction. Proceedings ofthe 2?h annual

I meeting ofthe North East Linguistic Society. 17-32. Amherst, MA: GLSA

Publications.

AnderBois, Scott. 20089. Strong positions and laryngeal features in Yucatek Maya.

1 Proceedings ofthe 39 h annual meeting ofthe North East Linguistic Society.

Amherst, MA: GLSA Publications.

Andersen, August, William Kalleo & Beatrice Watts. 2007. Labradorimi ulinnaisigutet:

An lnuktitut-English dictionary ofthe northern Labrador dialect. Nain, NL:

Tomgasok Cultural Centre.

Berinstein, Ava. 1979. A cross-linguistic study on the perception and production of stress.

UCLA working papers in phonetics 47. 1-59.

Bj0rnum, Stig. 2003. Gronlandsk grammatik. : Forlaget Atuagkat.

Bobaljik, Jonathan David. 1996. Assimilation in the Inuit languages and the place of the

uvular nasal. International journal ofAmerican linguistics 62(4): 323-350.

Bourquin, Theodor. 1891. Grammatik der Eskimo-sprache. London: Moravian mission

agency.

131 Bye, Patrik. 2011. Dissimilation. In Oostendrop, M. van, C. Ewen, E. Hume & K. Rice

(eds.), The Blackwell Companion to Phonology Vol. 2. 1285-1308. Oxford: Wiley­

Blackwell.

Chomsky, Noam & Morris Halle. 1968. The sound pattern ofEnglish. Cambridge, MA:

MIT Press.

Clements, George N. & Samuel Keyser. 1983. CVphonology: A generative theory ofthe

syllable. Cambridge, MA: MIT Press.

Collis, Dirrnid. 1970. Etudes philologiques et linguistiques des languages Esquimaux.

Inter nord Vol. 11. 263-282.

Creider, Chet. 1981. Place of articulation assimilation and the Inuktitut dialect continuum.

Etudes/Inuit/Studies 5(supplement). 91-1 00.

Dorais, Louis-Jacques. 1976. Some phonological characteristics of eastern Canadian

Inuit. Anthropological linguistics 18(8). 387-92.

Dorais, Louis-Jacques. 1977a. Les dialectes Inuit de !'arctique oriental Canadien: Une

comparison phonologique. Etudes/Inuit/studies 1(2). 4 7-56.

Dorais, Louis-Jacques. 1977b. Les parler Inuit de Rigolet (Labrador). Etudes/Inuitlstudies

1(2). 113-117.

Dorais, Louis-Jacques. 1990a. Inuit uqausingit 1, 000 mots Inuit. Kuujjuak, QC:

Association Inuuksiutiit Katimajiit.

Dorais, Louis-Jacques. 1990b. Inuit uqausiqatigiit: Inuit languages and dialects.

Kuujjuak, QC: Association Inuuksiutiit Katimajiit.

132 Dorais, Louis-Jacques. 2003. Inuit Uqausiqatigiit: Inuit languages and dialects, 2 nd

edition. lqaluit, NU: Nunavut Arctic College.

Dorais, Louis-Jacques & Ronald Lowe. 1982. Les dialectes de ['arctique de !'ouest.

Etudes/Inuitlstudies 6(2). 127-33.

Dresher, Elan & Alana Johns 1990. The law of double consonants in Inuktitut. Paper

1 presented at the 35 h annual meeting of the Canadian Linguistics Association,

Victoria, B.C.

Dresher, Elan & Alana Johns. 1995. The law of double consonants in Inuktitut.

Linguistica Atlantica Vol. 17. 79-95.

Dresher, Elan & Alana Johns. 1996. Rigolet and phonological change in Labrador.

Etudes/Inuit/studies 20( 1). 113-121 .

Egede, Hans. 1925. Relationer fra Gnmland 1721-36 & Det gamle Gmlands ny

perlustration 1741. In Bobe, L. (ed.), Meddelelser om Gronland 54. 1-304 & 305-

404. Copenhagen: Bianco Luno.

Egede, Poul. 1760. Dictionarium Gronlandica Danico-Latinum. Havnire.

Erdmann, Friedrich. 1864. Eskimoisches worterbuch gesammelt von missionaren in

Labrador. Bautzen: Ernst Moritz Monse.

Fortescue, Michael. 1983. A comparative manual of affixes for the Inuit dialects of

Greenland, Canada and Alaska. Meddelelser om Gronland 229. Copenhagen: Nyt

Nordisk Forlag.

Fry, Dennis B. 1958. Experiments in the perception of stress. Language and speech 1(2).

126-152.

133 Geytenbeek, Brian B. & Helen Geytenbeek. 1971. Gidabal grammar and dictionary.

Canberra, ACT: Australian Institute of Aboriginal Studies.

Gragg, Gene. 1976. Oromo ofWellegga. In Bender, M. L. (ed.), The non-Semitic

languages ofEthiopia. 166-195. East Lansing, MI: Michigan State University

Press.

Gussenhoven, Carlos. 2000. The boundary tones are coming: On the non-peripheral

realization of boundary tones. In Broe, M. & J. Pierrehumbert (eds.), Papers in

laboratory phonology acquisition and the lexicon. 132-152. Cambridge: Cambridge

University Press.

Goldsmith, John. 1975. Tone melodies and the autosegment. In Herbert, R. (ed.),

Conference onAfricanLinguistics. 135-147. Columbus, OH: Ohio State University.

Goldsmith, John. 1976. Autosegmental phonology. Cambridge, MA: MIT doctoral

dissertation.

Hagerup, Asger. 2011. A phonological analysis ofvowe l allophony in West Greenlandic.

Bergen, Norway: Norwegian University of Science and Technology MA thesis.

Harris, John. 1994. English sound structure. Cambridge, MA: Blackwell.

Hayes, Bruce. 1985. Iambic and trochaic rhythm in stress rules. Proceedings ofthe IF"

Annual Meeting ofthe Berkeley Linguistics Society. 429-446. Berkeley, CA: La

Revue de Sommaires.

Hayes, Bruce. 1987 A revised parametric metrical theory. Proceedings of the 17111 annual

meeting of the North East Linguistic Society. 274-289. Amherst, MA: GLSA

Publications.

134 Hayes, Bruce. 1989. Compensatory lengthening in moraic phonology. Linguistic inquiry

20(2). 253-306.

Hayes, Bruce. 1995. Metrical stress theory: Principles and case studies. Chicago, IL:

UCP.

Iverson, Gregory, & Joseph C. Salmons. 1995. Aspiration and laryngeal representation in

Germanic. Phonology 12(3). 369-96.

Ito, Junko. 1989. A prosodic theory of epenthesis. Natura/language & linguistic theory

7(2). 217-259.

Ito, Junko & Armin Mester. 1998. Markedness and word structure: OCP effects in

Japanese. University of California manuscript.

Iwai , Melissa. 1989. A Prosodic analysis ofJapanese loanwords. Santa Cruz, CA:

University of California BA thesis.

Jacobsen, Birgitte. 2000. The question of 'stress ' in West Greenlandic: An acoustic

investigation ofrhythmicization, intonation, and syllable weight. Basel: S. Karger

AG.

Jeddore, Rose. 1976. Labrador Inuit uKauset katitsutauppaliningit. St. John's, NL:

Labrador Inuit Committee on Literacy.

Kager, Rene. 1993. Alternatives to the iambic-trochaic law. Natura/language &

linguistic theory 11(3). 381-432.

Kager, Rene. 1995. Review of metrical stress theory: Principles and case studies.

Phonology 12(3). 437-464.

135 Kahn, Daniel. 1976. Syllable-based generalizations in English phonology. Bloomington,

IN: Indiana University Linguistics Club.

Kaplan, Lawrence. 1979. Phonological issues in North Alaskan Inupiaq. San Diego, CA:

University of California San Diego doctoral dissertation.

Kaye, Johnathan. 1990. Coda licensing. Phonology 7. 301-330.

Kenstowicz, Michael J. & Charles Kisseberth. 1977. Topics in phonological theory. New

York, NY: Academic Press.

Kenstowicz, Michael J. & Charles Kisseberth. 1979. Generative phonology: Description

and theory. New York, NY: Academic Press.

Keyser, Samuel & Paul Kiparsky. 1984. Syllable structure in Finnish phonology. In

Aronoff, M. & R. Oehrle (eds.), Language sound structure. 234-244. Cambridge,

MA: MIT Press.

Kleinschmidt, Samuel. 1851. Grammatik der Gronlandischen sprache mit teilweisem

einschluss des Labradordialekts. Hildesheim: Olms.

Kleinschmidt, Samuel & Theodor Bourquin 1881. The letters ofSamuel P Kleinschmidt

to Theodor Bourquin, August 1865-May 1881. In Schloss, B. (trans.) ofHoltved, E.

(ed.), Kleischmidts briefe an Theodor Bourquin. Copenhagen: C. A. Reitzels.

Krupnik, Igor & Winton Weyapuk, Jr. 2010. Qanuq ilitaavut, how we learned what we

know: Wales Inupiaq sea ice dictionary. In Krupnik, 1., C. Aporta, S. Gearheard, G.

Laidler & L. Kielsen Holm (eds.), SIKU: Documenting Inuit sea ice knowledge and

use. 321-354. Dordrecht: Springer.

Ladefoged, Peter. 1975. A course in phonetics. New York: Harcourt Brace Jovanovich.

136 Leumann, Manu. 1977. Lateinische grammatik von Leumann-Hofmann-Szantyr Vol. 1.

Lateinische Laut- und Pormenlehre. Munich: Beck.

Lieberman, Mark & Alan Prince. 1977. On stress and linguistic rhythm. Linguistic

inquiry 8(2). 249-336.

Lloret, Maria-Rosa. 1988. Gemination and vowel length in Oromo .

Indiana, IN: Indiana University doctoral dissertation.

Lowe, Ronald. 1984. Basic Siglit Inuvialuit Eskimo dictionary. Inuvik, QC: Association

Inuksuit Katimajiit Inc.

Lowe, Ronald. 1985. Basic Siglit Inuvialuit Eskimo grammar. Inuvik, QC: Association

Inuksuit Katimajiit Inc.

Mase, Hideo. & J0rgen Richel. 1971. A study of consonant quantity in West Greenlandic.

1 University ofCopenhagen 5 h annual report ofthe institute ofphonetics. 175-247.

Mase, Hideo. 1973. A study of the role of syllable and mora for the tonal manifestation in

West Greenlandic. University ofCopenhagen Th annual report ofthe institute of

phonetics. 1-98.

Massenet, Jean-Marie. 1980. Notes sur !'intonation finale en Inuktitut. Cahiers de

linguistique 10. 195-214.

Massenet, Jean-Marie. 1986. Etudes phonologiques d 'un dialecte Inuit Canadien. Inuvik,

QC: Association Inuksuit Katimajiit Inc.

137 Mehler, Jaques, Marina Nespor and Mohinish Shukla. 2011. Stress-timed and syllable­

timed languages. In Oostendorp, M. van, C. J. Ewen, E. Hume & K. Rice (eds.),

The Blackwell companion to phonology. 1147-1159. Hoboken, NJ: Wiley­

Blackwell.

McCarthy, John & Alan Prince. 1986. Prosodic morphology. University of Massachusetts

and Brandeis University manuscript.

Nagano-Madsen, Yasuko. 1988. Phonetic reality of the mora in West Greenlandic Eskimo.

Lund University working papers 34. 79-82.

Nagano-Madsen, Yasuko. 1990. Quantity manifestation and mora in West Greenlandic

Eskimo: Preliminary analysis. Lund University working papers 36. 123-132.

Nagano-Madsen, Yasuko. 1993. Phrase-final intonation in West Greenlandic Eskimo.

Lund University working papers 40. 145-155.

Nagano-Madsen, Yasuko. 1994. Mora and prosodic coordination: A phonetic study of

Japanese, Eskimo and Yoruba. Lund: Lund University Press.

Nagano-Madsen, Yasuko & Anne-Christine Bredvad-Jensen. 1995. Analysis of

intonational phrasing in West Greenlandic Eskimo reading text. Lund University

working papers 44. 129-144.

Nespor, Marina & Irene Vogel. 1986. Prosodic phonology. Dordrecht: Foris.

Ramus, Franck, Marina Nespor & Jacques Mehler. 1999. Correlates of linguistic rhythm

in the speech signal. Cognition 73(3). 265- 292.

Rasmussen, Jens. 1979. Anaptyxis, gemination and syncope in Eskimo, a diachronic

study. Copenhagen: The linguistic circle of Copenhagen.

138 Reuse, Willem de. 1994. Siberian Yupik Eskimo: The language and its contacts with

Chukchi. Salt Lake City, UT: University of Utah Press.

Rischel, Jorgen. 1974. Topics in West Greenlandic phonology. Copenhagen: Akademisk

Forlag.

Roach, Peter. 1982. On the distinction between 'stress-timed' and 'syllable-timed'

languages. In Crystal, D. (ed.), Linguistic controversies. 73-79. London: Edward

Arnold.

Piggott, Glyne. 1998. Foot form and the parsing of weightless syllables. In Gruber, M., D.

Higgins, K. Olson & T. Wysocki, (eds.). Papers from the 34111 regional meeting of

the Chicago Linguistic Society. 315-332. Chicago, IL: Chicago Linguistics Society.

Prince, Alan. 1984. Phonology with tiers. In Aronoff, M. & R. Oehrle (eds.), Language

sound structure. 234-244. Cambridge, MA: MIT Press.

Pyle, Charles. 1970. West Greenlandic Eskimo and the representation ofvowellength.

Papers in Linguistics 3(1). 115-146.

Rose, Yvan, Paul Pigott, & Douglas Wharram. 2012. Schneider's Law revisited: The

syllable-level remnant of an older metrical rule. In McKillen, A. & J. Loughran

( eds. ), Proceedings from phonology in the 21st Century: In honour ofGlyne

Piggott. Montreal, QC: McGill working papers in linguistics 22(1), Winter 2012. 12

pages.

Rubach, Jerzy. 1993. The lexical phonology ofSlovak. Oxford: Clarendon Press.

Schneider, Lucien. 1966. Grammarian Esquimaude du sous-dialecte de l 'Ungava.

Quebec, QC: Ministere des Richesses Naturelles.

139 Schneider, Lucien. 1972. Jnuktituorutit: Grammaire purement Esquimaude. Quebec, QC:

Ministere des Richesses Naturelles.

Schneider, Lucien. 1985 Ulirnaisigutiit/ An Inuktitut-English dictionary ofnorthern

Quebec, Labrador and eastern Arctic Dialects. Quebec, QC: Presses de l'Universite

Laval. (Originally published in French in 1970).

Selkirk, Elizabeth. 1980. On prosodic structure and its relation to syntactic structure.

Bloomington, IN: Indiana University Linguistics Club.

Sihler, Andrew L. 1995. New comparative grammar ofGreek and Latin. Oxford: Oxford

University Press.

Smith, Lawrence. 1975. Labrador Inuttut surface phonology. International journal of

American linguistics 41(2). 97-105.

Smith, Lawrence. 1977a. Some morphophonemic processes of Labrador Inuktitut

affixation. International journal ofAmerican linguistics 43(2). 77-84.

Smith, Lawrence. 1977b. Some grammatical aspects ofLab rador Inuttut (Eskimo): A

survey ofthe inflectional paradigms ofnouns and . Ottawa, ON: National

Museum of Man mercury series.

Smith, Lawrence. 1978. A survey ofderivational post bases ofLabrador Inuttut (Eskimo).

Ottawa, ON: National Museum of Man mercury series.

Smith, Lawrence. 1979. Some recent phonological and orthographic changes in Labrador

Inuttut. In Basse, B. & K. Jensen (eds.), Symposium on majority language influence

on Eskimo minority languages. I 08-111. Aarhus: Aarhus University.

140 Suzuki, Keiichiro. 1998. A typological investigation of dissimilation. Tuscon, AZ:

University of Arizona doctoral dissertation.

Tranel, Bernard. 1991. CVC light syllables, geminates and moraic theory. Phonology

8(2). 291-302.

Ulving, Tor. 1953. Consonant gradation in Eskimo. International journal ofA merican

linguistics 19( 1). 45-52.

Wade, Juliette. 1996. An examination ofword borrowings into Japanese. Santa Cruz, CA:

University of California MAf thesis.

Willis, Anne-Marie. 1971. Some phonological problems ofan Eskimo dialect ofthe

Ungava region. Ottawa, ON: University of Ottawa MA thesis.

Windsor, Joseph & Cobler, Jessi. 2013. A unified analysis of three phrase-final

phenomena in Blackfoot. Poster presentation at the 58 111 Annual meeting ofthe

Canadian Linguistics Association, Victoria, B.C.

Woodrow, Herbert. 1951. Time perception. In Stevens, S. S. (ed.), Handbook of

experimental psychology. 1224-1236. New York: Wiley.

141 APPENDIX A: Linguistic interview

A: Introductory questions 1. kinauven? I What is your name? 2. KatsinikjariKaven? I How old are you? 3. Nanemiunguven? I Where are you from? 4. Ananait, atataitilu, nanemiunguvang? I Where are you mother and father from? 5. Nane angiKait manna? Piujong? I Where are you living now? Is is good? 6. sunauna nigikKauven ullumi? I What have you eaten today? 7. PinasualautsimalaukKen puijet? I Have you ever hunted seals before? 8. UKalautiKalaulaget Kanong tuKilaukKan puijet sivullipak. I Tell the story about how you killed your first seal.

B: Introductory reading task Have the language consultant read Beatrice Watt's Inuttut introduction to the 2006 dictionary.

C: Phonemic pair reading task 1. Tanna aggak piujuk 2. Angutik ipiunnguangittuk ullumi 3. Angutik tutonnguangittuk ullurni 4. Maggonik anak piujok 5. Angutik ikittonguanngituk ullumi 6. Angutik taKaunnguangittuk ullurni 7. Angutik pisiunnguangittuk ullurni 8. Angutik niliunnguangittuk ullumi 9. Tanna itivinik piujuk 10. Angutik inngitiunnguangittuk ullumi 11. Angutik tingijonnguanianngituk ullumi 12. Tanna anak piujuk 13. Angutik aggaunguajuk ullumi

142 14. Angutik ikketujonnguangittuk ullumi 15. Angutik aggaunguangituk ullumi 16. Angutik kiviniunnguangittuk ullumi 1 7. Angutik iginaunnguangittuk ullumi 18. Angutik imennguangittuk ullumi 19. Angutik tingijonnguangittuk ullumi 20. Angutik ijiunnguangittuk ullumi 21. Angutik alaunnguangittuk ullumi 22. Angutik ippiunguanngituk ullumi 23. Angutik tuttonguanngituk ullumi 24. Angutik KakKanguanngituk ullumi 25. Angutik pitsiunguanngituk ullumi 26. Angutik nilliunguanngituk ullumi 27. Angutik annaunguanngituk ullumi 28. Angutik mipviunguanngituk ullumi 29. Angutik aKiggiunguanngituk ullumi 30. Angutik Kimmiunguanngituk ullumi 3 1. Angutik itjiliunnguangittuk ullumi 32. Angutik atlaunguanngituk ullumi 33. Angutik anannguangittuk ullumi 34. Angutik amlnnguangittuk ullumi 35. Angutik innenguanngituk ullumi 36. Angutik itennguangittuk ullumi 37. Angutik ullonguanngituk ullumi 38. Angutik ojonnguangittuk ullumi 39. Maggonik anak piujok 40. Tanna enniuvinik piujuk 41. Angutik anaunnguajuk ullumi 42. Angutik ojonnguanianngituk ullumi 43. Angutik anaunguanngituk ullumi

143 44. Tanna innik piujuk 45. Angutik aggaunguanianngituk ullumi 46. Angutik ullonguanianngituk ullumi 47. Tanna ikketujovinik piujuk 48. Angutik ujugonnguajuk ullumi 49. Angutik inniunguajuk ullumi 50. Angutik inniunguanngituk ullumi 51. Angutik anaunnguanianngituk ullumi 52. Tanna ikik piujuk 53. Angutik ullonguajuk ullumi 54. Angutik agganguanngituk ullumi 55. Maggonik innek piujok 56. Angutik iginaunnguajuk ullumi 57. Tanna ulluk piujuk 58. Angutik ikkiunguanngituk ullumi 59. Tanna ujuguk piujuk 60. Angutik ullonguanngituk ullumi 61. Tanna aggak piujuk 62. Angutik inniunguanianngituk ullumi 63. Angutik ikkiunguanianngituk ullumi 64. Angutik ullonguanianngituk ullumi 65. Angutik ujugonnguanianngituk ullumi 66. Angutik agganguanianngituk ullumi 67. Tanna anak piujuk 68. Angutik enniunguanianngituk ullumi 69. Angutik ikketujonnguanianngituk ullumi 70. Angutik ullonguanianngituk ullumi 71. Angutik ojunnguanianngituk ullumi 72. Tanna ennik piujuk 73. Angutik imiunnguanianngituk ullumi

144 74. Angutik ijiunnguanianngituk ullumi 75. Angutik alaunnguanianngituk ullumi 76. Angutik ippiunguanianngituk ullumi 77. Taitsumani tutunnguaKalluni Nainimelauttuk 78. Angutik tuttonguanianngituk ullumi 79. Angutik KakKaunguanianngituk ullumi 80. Angutik pitsiunguanianngituk ullumi 81. Tanna ikketujok piujuk 82. Tanna ullok piujuk 83 . Angutik ikkinguajuk ullumi 84. Tanna ojuk piujuk 85.Tanna itik piujuk 86. Angutik agganguajuk ullumi 87. Tanna ipik piujuk 88. Tanna ikittut piujuk 89. Angutik ujugunnguangittuk ullumi 90. Angutik enniunguanngituk ullumi 91. Tanna tutuk piujuk 92. Angutik ullonguanngituk ullumi 93. Angutik ojonnguangittuk ullumi 94. Tanna taKak piujuk 95. Angutik ikketujonnguajuk ullumi 96. Tanna pisik piujuk 97. Angutik nilliunguanianngituk ullumi 98. Angutik annaunguanianngituk ullumi 99. Angutik mipviunguanianngituk ullumi 100. Angutik aKiggiunguanianngituk ullumi 101. Angutik Kimmiunguanianngituk ullumi 102. Angutik inngitinnguanianngituk ullumi 103. Angutik itjiliunnguanianngituk ullumi 104. Angutik atlaunguanianngituk ullumi 105. Taitsumani ikketujonnguaKalluni Nainimelauttuk

145 106. Angutik amlnnguanianngituk ullumi 107. Angutik anannguanianngituk ullumi 108. Angutik innenguanianngituk ullumi 109. Angutik itennguanianngituk ullumi 110. Tanna nilik piujuk 111. Angutik enniunguajuk ullumi 112. Tanna kivinik piujuk 113. Angutik ullonguajuk ullumi 114. Tanna iginak piujuk 115. Angutik ojunnguajuk ullumi 116. Tanna imik piujuk 117. Angutik ipiunnguanianngituk ullumi 118. Angutik tutonnguanianngituk ullumi 119. Angutik ikittonguanianngituk ullumi 120. Angutik ippiunguajuk ullumi 121. Maggonik itek piujok 122. Angutik taKaunnguanianngituk ullumi 123. Angutik pisiunnguanianngituk ullumi 124. Angutik niliunnguanianngituk ullumi 125. Angutik kiviniunnguanianngituk ullumi 126. Angutik iginaunnguanianngituk ullumi 127. Tanna tingijok piujuk 128. Angutik mipviunguajuk ullumi 129. Tanna ijik piujuk 130. Angutik annaunguajuk ullumi 131. Tanna alak piujuk 132. Angutik nillinguajuk ullumi 133. Angutik aggaunguakKaungittuk ullumi 134. Angutik anannguaKaunngituk ullumi 135. Angutik inninguakKaungittuk ullumi

146 136. Angutik ikkinguakKaungittuk ullumi 137. Angutik ullunguakKaungittuk ullumi 138. Angutik ujugonnguaKaunngituk ullumi 139. Angutik agganguakKaungittuk ullumi 140. Angutik enniunguakKaungittuk 141 . Angutik ikketujonnguaKaunngituk ullumi 142. Taitsumani annanguakK§.luni Nainimelauttuk 143. Angutik ullonguakKaungittuk ullumi 144. Angutik ojonnguaKaunngituk ullumi 145. Angutik ipiunnguaKaunngituk ullumi 146. Angutik tutonnguaKaunngituk ullumi 147. Angutik ikittonguakKaungittuk ullumi 148. Angutik taKaunnguaKaunngituk ullumi 149. Angutik pisiunnguaKaunngituk ullumi 150. Angutik niliunnguaKaunngituk ullumi 151. Taitsumani nanunnguaKalluni Nainimelauttuk 152. Angutik kiviniunnguaKaunngituk ullumi 153. Angutik iginaunnguaKaunngituk ullumi 154. Angutik imiunnguaKaunngituk ullumi 155. Angutik tingijonnguaKaunngituk ullumi 156. Taitsumani enninguakKaluni N ainimelauttuk 15 7. Angutik ij i unnguaKaunngi tuk ull umi 158. Angutik alaunnguaKaunngituk ullumi 159. Angutik ippiunguakKaungittuk ullumi 160. Angutik tuttonguakKaungittuk ullumi 161. Angutik KakKaunguakKaungittuk ullumi 162. Angutik pitsiunguakKaungittuk ullumi 163. Angutik nilliunguakKaungittuk ullumi 164. Angutik annaunguaKaunngituk ullumi 165. Angutik mipviunguakKaungittuk ullumi

147 166. Angutik aKiggiunguakKaungittuk ullumi 167. Taitsumani nillinguakKaluni Nainimelauttuk 168. Angutik KimmiunguakKaungittuk ullumi 169. Angutik inngitiunnguaKaunngituk ullumi 170. Angutik itjiliunnguaKaunngituk ullumi 171. Angutik atlaunguakKaungittuk ullumi 172. Angutik anannguaKaunngituk ullumi 173. Angutik anannguaKaunngituk ullumi 174. Angutik innenguakKaungittuk ullumi 175. Angutik itennguaKaunngituk ullumi 176. Angutik ullonguakKaungittuk ullumi 177. Angutik ojonnguaKaunngituk ullumi 178. Tanna ippik piujuk 179. Angutik aKiggiunguajuk ullumi 180. Angutik inngitinnguajuk ullumi 181. Tanna tuttuk piujuk 182. Angutik alaunnguajuk ullumi 183. Tanna ikkik piujuk 184. Angutik tuttonguajuk ullumi 185. Angutik ijiunnguajuk ullumi 186. Maggonik ullok piujok 187. Angutik KakKaunguajuk ullumi 188. Angutik tingijonnguajuk ullumi 189. Tanna KakKak piujuk 190. Angutik ipiunnguajuk ullumi 191. Tanna pitsik piujuk 192. Angutik tutonnguajuk ullumi 193. Tanna nillik piujuk 194. Angutik imiunnguajuk ullumi 195. Tanna annak piujuk

148 196. Angutik ullonguajok ullumi 197. Tfuma mipvik piujuk 198. Angutik ikittonguajuk ullumi 199. Tfuma aKiggik piujuk 200. Angutik taKaunnguajuk ullumi 201. Tfuma Kimmik piujuk 202. Angutik pisiunnguajuk ullumi 203. Tanna inngitit piujuk 204. Angutik pitsiunguajuk ullumi 205. Maggonik ojok piujok 206. Angutik Kimmiunguajuk ullurni 207. Tanna itjilik piujuk 208. Angutik niliunnguajuk ullurni 209. Angutik itennguajuk ullurni 210. Tanna atlak piujuk 211. Angutik kiviniunnguajuk ullurni 212. Tfuma ojuvinik piujuk 213. Angutik itjiliunnguajuk ullurni 214. Tfuma tuttuvinik piujuk 215. Angutik atlaunguajuk ullumi 216. Tanna nillivinik piujuk 217. Angutik anfumguajuk ullurni 218. Tanna aKiggivinik piujuk 219. Angutik anannguajuk ullumi 220. Tanna Kimmivinik piujuk 221. Angutik innenguajuk ullurni 222. Tfuma atlavinik piujuk 223. Angutik ojonnguajuk ullumi

149 D: tutuk/tuttuk alternation task. Through translator, ask the language consultant to use these words in a conversational sentence, without showing them the words or using the following constructions: 224. Tuttonguanianngituk 225. Tutonnguanianngituk 226. TuttonguakK.aungittuk 227. TutonnguaKaunngituk

E: Story telling. Ask the following questions through translator: 1. What do you remember about the time Gus Bennett, Martin Sillit and Paul Semigak died caribou hunting in that blizzard of 1979 and what do you think happened? 2. What changes have you notice in the weather since the 1979, and what about since you were young? 3. What is different about the ice conditions? What changes have you seen? 4. What are some of the words your Grandparents used that you don't hear often today?

150 APPENDIX B: Ethical Approval

Certification oflnformed Consent (as approved by ICEHR) July 28, 2009

ICEHR No. 2008/09-072-AR Mr. Paul Pigott Department of Linguistics Memorial University ofNewfoundland

Dear Mr. Pigott:

Thank you for your correspondence of April 14, 2009, addressing the issues raised by the Interdisciplinary Committee on Ethics in Human Research (ICEHR) concerning your research proposal "Inuttut ice topology".

We are happy to confirm our earlier approval of your proposal. If you intend to make changes during the course of the project which may give rise to ethical concerns, please forward a description of these changes to the ICEHR Co-ordinator, Mrs. Eleanor Butler, at [email protected] for the Committee's consideration. The Tri-Council Policy Statement on Ethical Conduct for Research Involving Humans (TCPS) requires that you submit an annual status report on your project to ICEHR, should the research carry on beyond March 2010. Also, to comply with the TCPS, please notify us when research on this project concludes. We wish you success with your research. Yours sincerely, Lawrence F. Felt, PhD

Chair, Interdisciplinary Committee on Ethics in Human Research

LF/bl

151 APPENDIX C: The phonemic pair /tutuk/ 'messy hair' versus /tuttuk/ 'caribou'

Derivational ordering (see Dresher & Johns 1995:83)

Truncation= stem-final consonants are deleted before suffixation

RPA = regressive place assimilation

SL = applies after Truncation and Assimilation cycles

Affrication = voiceless spirant geminates are affricated l XXI ---+ [qx], I ss/ ---+ [ts], / 11/ ---+ [tl], !jj! ---+ [dz]

Light grey = analyzed as an underlying geminate, or SL trigger

ms = segmental duration values averaged to the nearest 0.001 second

0.97s = measurement of the sequence / tutu:IJIJUaxaUIJIJitu-/ = peak ffi intensity

h = phrase/utterance-fi nal aspiration

152 (1) /tutu(k) + u + DDUa + xxau + DDi(k) + tuk/ UR (2) /tuttu(k) + u + DDUa + xxau + DDi(k) + tuk/ messy hair. to be.pretend.near past.negative.3s Gloss caribou. to be.pretend.near past.negative.3s tutu:DDuaxxaUDDiktuk Truncation tuttu:DDUaXXaUDDiktuk tutu:DDuaxxauDDittuk RPA tuttu:DDUaXXaUDDittuk tu tu:DDuaxa UDDi tuk SL Affrication [tutu:DDuax aUDDi tuk] SR Key: Sequence Segment Final 5e a. t u It u: i It u k" SL t rigger SL target Segment duration Du ra t ion 28 28 169 121 41 114 28 489m5 Seq dur Position 3.9 3. 5 e. 96 a. t u t u: 'au q i It u kh Surface b. t u It u ua X 'au qq i u 43 18 91 126 197 126 189 88 42 169 43 381m5 TK 36 93 139 49 237 183 85 31 66m5 1.3 6. 975 2. 1 1.3 165 b. t u t u: qq 'ua u k c. t u It u: i It u k 28 61 67 111 168 124 95 76m5 BH 21 38 78 125 27 56m5 1.2 SL 1.5 1675 1.4 1.19 c. t u t u qq ua 'au q i It u k d. t u It u 'ua X au q i: t u k" 62 69 88 22 289 172 93 124 45 128 23 188m5 MK 19 88 285 87 227 63 127 88 129 21 91 149ms 1.5 1.165 2. 6 1.22 d. t u t u: qq 'ua au q i It u kh e. t u It u i It u kh w 26 37 126 128 188 186 98 124 127 34 121 28 382m5 MH 47 181 148 181 74 191 188 153m5 2 1. 285 1.6 1.29 e. t u t u: qq 'ua au q i It u kh f. t u It 'u: t u k" 29 49 122 163 149 191 137 111 31 139 32 165m5 MH 64 58 258 198 193 88 34 389m5 1.5 1.275 2. 7 1. 44 f. t u t u: qq 'ua au i It u th g. t u It u: t u k" 44 128 61 153 158 166 1 184 57 82 117 68 263m5 FW 89 199 244 199 198 58 199 23 4m5 1.2 1.3 35 3. 6 1.53 g. t u i It u kh h. t u It tu: It u 37 147 47 149 165 248 46 249 137 58m5 JD 17 76 61 28 5 2. 3 1.395 2. 7 2. 27 h. t u t u: qq 'ua au i It u k 54 122 59 153 91 52 249 88 11lm5 PA 1.5 1. 455 ~ 40 82 154 102 99 147 Segment 42 212 85 146 71 134 1.23 5 Sequence 1.37 ~ 1.6 1.5 Gem Short 2. 6 2. 7 1.3 [tl avg = 82 [tl: [ttl = 1. 9 c vs. cc [tl avg = 69 [t]: [ttl = 3 [ttl avg = 156 [ttl avg = 288 [Q] avg = 77 [Q] : [QQ ]=l.9 [Q Q]= 143

153 (3)/tutu(k) +u +DDUa +nia(x) + DDi(k) + tuk/ UR (4) /tuttu(k) +u +DDUa +nia(x) +DDi(k) +tuk / messy hair. to be. pretend. near future.negative.3s Gloss caribou. to be.pretend.near future.negative.3s tutu:DDUaniaDDiktuk Truncation tuttu:DDUaniaDDiktuk tutu:DDUaniaDDittuk RPA tuttu:DDUani~Di t tuk tutu:DDUaniaDDituk SL tuttu:Duani~Dituk Atfrication

__,------'[,_tu_tu_,:D_Duani~Q_ituk~],__-----,----1 SR 1--r----[ tuttu:Duani~D_itu~k ],____---.--__ a. t u t 'u: qq ua n ia q i t u k Surface a. t u tt u: fij\i 'ua ~ -D i t u kh 24 39 184 113 119 120 65 93 102 31 111 46 38m5 TK 39 28 164 164 6~ 117 68 79 96 47 95 45 365m5 TK 0. 975 2 8. 965 b. t u t u: qq 'ua n ia qq i t11 u kh b. t u t t u ua n 'ia Q i t u kh 29 75 52 128 114 161 63 126 111 99 ~ 97 125 m5 BH 96 39 184 68 183 76 131 111 62 129 29 818m5 MK 1.9 1. 85 s 3.9 1.155 c. t u t 'u qq ua n ia qq i !1 u k c. t u tt u: k 28 83 84 78 195 153 98 126 136 58 :~ 39 25m5 MH 52 67 188 146 128 73 128 124 99 5 43 183m5 BH 1.3 1.125 2.2 1.155 d. t u t u qq 'ua n ia q i t u k 26 89 113 96 171154 98 123 117 38 128 33 368m5 MH 27 93 127 68 · 184 73 123 115 58 115 63 173m5 PA 1.145 1.155 e. t u t u QQ 'ua n ia Q j t u kh k

72 66 46 67 154 187 53 176 87 97 65 114 146m5 PA 48 96m5 FW 1.175 1.4 1.385 f. t u t u qq 'ua n ia qq i: ~ u kh f.tuau: 'uaniaQit u k iPml 67 86 73 34 168 178 165 149 189 mm52 766m5 MK 34 117 173 138 183 87 147 84 61 83 56 154m5 JD 3 1.195 1.5 1.345 g. t u t u: qq 'ua n ia Q j t U· ua n 'ia Q i t u kb 28 46 97 113 149 178 87 139 79 86 49 184 m5 JD 173 96 158 118 51 129 32 279m5 MH 1. 285 1.9 1. 415 h. t u t 'u DD ua n ia D j tt u: xh h. t u tt 'u: 68 97 39 13 164 182 98 111 86 76 m 258 553ms MK 68 76 291 198 1.315 2. 4 Averages 43 76 153 88 76 83 Segment 51 185 91 76 86 97 1.165 Sequence 1.255 Ratios 2.1 SL 2. 5 1.9 [q]: [qq] = 1.6 cvs. cc [t]: [ttl = 2.1

154 Appendix D: The morpheme /xattax/ 'often, intermittently'

(5)a. k a tt u i v a ll ia ui nn a X a tt a u 142 99 174 70 84 60 51 179 138 101131 71 37 33 172 64 61 0.3ls b. i v u ll iu X a tt a a 87 66 77 49 182 149 65 30 171 63 37 0. 33s c. X ai DD u: y a tt au X a tt a u:h 132 122 187 127 45 52 123 211 148 20 22 299 78 296 0.33s d. a tj y i ts ia X a tt a t u: k 150 147 109 49 96 241 137 39 44 218 57 72 81 190 38 9. 34s e. s a: tt au X a tt a u: 77 128 287 158 28 48 192 182 166 0.375 f. p au X a tt a u: 47 68 101 132 49 40 296 84 154 0.375 g. i k ua X a tt a u: 166 52 153 174 42 54 199 83 276 0.375 h. X ua y u nn a iu m a X a tt a u th 186 204 41 59 189 68 140 93 47 53 38 219 73 43 147 0. 375 I. p X a tt a u t 35 69 22 65 223 67 88 39 9. 38s

155 For each example word the decibel value boxed in double lines attains peak word intensity. Decibel value boxed valued in a single line attains peak intensity for the [xatta] sequence.

mm au i I ua n X a tt a 113 152 187 57 91 158 78 136 66 53 181 79 295 9.385 k. X au s m a u: j a: X a tt a 162 84 49 95 89 67 289 53 155 43 52 296 99 784 8.395 I. X n ua IJ u n ya:x atta v u t 166 69 174 162 68 91 64 65 68 78 8.395

m. X n ua x a tt 228 88 192 248 55 74 69 67 217 55 185 9.485 n. a tj i: x a tt a 148 192 281 84 53 239 63 8.435 o.l uxattal)a 148 77 86 226 75 187 33 8.465 p. n X a tt a s u n i: 86 79 85 72 75 58 227 118 74 64 172 281 8.475 q. u xa ttatiun 183 199 67 182 62 219 182 74 97 172 9.485 r. X u pp a x a tt a u: y a I ua 97 188 219 98 178 96 299 84 199 24 65 55 385 8.565 68 52 285 89 88 8.495

156 (6)a. [ X a nn i t au s i m a j u: j a: x a tt _a_ u:h 62.6([]67.3 68.5 63.9 71.1 60 64.163.8 66.9 64.5 69.4 66.6 71.163 .9 62.2 59 .5[63.9 57 .8 63.5 58 .8 b.

c.

d. X ua y u nn a g iu m a X a tt a t u th 63.8~68.9 71 .5 71.8 75 .1 71 .7 72.4 69.5 71.5 64 63.6 69 .3 67.8 62 64.4 52 .8 61.1 e. t i n i x a tt a s u n i: 59. 4[LL]70.2 70.2 64 .5 65.7 58. 7 67. 8 56.3 65.3 65 .6 65. 7 61.4 f. u x a tt a D a I 67.9 58 .7 ~61.6 68.5 65 . 8 ~ 61.8 g. i mm au i I w n i X a tt a ~ 65.2 64.s!W] 66 .9 65 .4 ~65 . 5 71.2 68. 4 63 .5 64 .7[68.8 63.8 54 .9 62 .1 h. k a tt u X i v a 11 ia t ui nn a x a tt a t u 60.9 70. 7 63.86]65.5 72.3 67.8 71.5 68.5 73 .4 61.7 70 .2 68 .7 69 .2 65 .9 63 58 ~ 64 .165. 4 62 .7 l. X ai DD u: n i y a tt au x a tt a u:h 65 . 4~71.2 72.8 70.8 71 .6 68.8 73.6 64.5 70 .4 67.9 64.9 62 69.3 59 .6 66. 8 63

157 J. s a: tt u: 61.3 71.5 63 ' 63 .1 k. u l i X a tt a ri u n lil170 ' 1 71.6 6U 68 .1 60.5 69 .6 6U 65 .6 6U 63 .2

I. X I n ua l I X a tt a a 64.5 7U 69 . 1~66 . 6 69 . 6 5 . 7 ~ 64.2 64 .5 65 .3 68.2 6U m. a g i y i ts ia t u: k 71 s7.9lZuhu 71.1 67 72.2 .8 63 .5 60.2 71 63 .4 n. I tt a

0. X i n ua D u n i y a: a v u t 67 .6 73 73.4 ~ 73. 1 73.8 71.8 72.8 70.4 76 p. X u pp a X a tt a u: -y a l ua 58 .5 71.3 68 .4Ei]65.7 70 .8 66.6 73.7 69 .251]66.3 72.2 65. 4 q. v u ll 59 ' 9 69' 5 65 ' 5 70' 1 69 '

r. p X a tt a u t .9 ~ 71.5 71. 4 66.9 70.4 61.4 68 .6 Avg. 65.7 66.2 62 .2 68 .7 6U

158 APPENDIX E: The sequence /xixxi/

For all the tables in the ptarmigan corpus, peak intensity is shown above each segment in decibels with the peak word intensity value boxed in double lines and secondary word intensity boxed in a single line. For duration, values are beneath each segment in milliseconds with peak word duration boxed in double lines and secondary durational prominence boxed in a single line. The examples are ordered from the shortest [xixxil sequence in seconds. The boundary tones H and L are marked with the unmarked case being a pitch line slightly falling from left to right. Prosodic boundaries are marked, as needed for my analysis. ]w= word boundary ]Jr= interrogative phrase-final boundary ]or = declarative phrase-final boundary ]u = utterance-final boundary

159 71 61 63 56 ~dB 7a. ~ X i x I x ~ ] "' BH AE 10 4 2 5 9 0 57 57 ll117llms 0 . 355 I

61 61 59 117511 68 64 65 ~9 70 1 b 0 X i X I X i ] BH X 1- y I y -: k h ]DP HW 14 56 91 91 1111611 w 81 93 61 6111214 11118 0.365 1 0.515

72 ~ 64 69 52 X i y I y t:i kh ]IP]U HP 49 96 76 76112541/292 0 . 555

57 57 66 60 11541146 1 4_8 1 44 ~3 ~ 3~ d. a x i x x t:i ] P]u JM 1 j.=9 X 1 X I X 1: k ]DP AZ [1}l 79 7 5 109 1091148911 8 3 3 6 6 7 91 91112 6 0113 4 2 0 . 865 0.555 I

64 65 ~9 70 f X I X i: kh ]DP BK 58 96 9612241343 0.595

56 58 6 ~ 60 64 56 f. a X i x I x 1: I kh 1u A E 64 170166] 114 1141112811224 0.495 I

160 67 62 66 62 i 71 157 . , kh ] BH g. a x 1 x x 1: u AE 96 112 140 63 163 124 423 96 315 II 610 II 64 65 ry] 78 h. a X y y i: kh ] DP HW 111 81 93 61 ,61 214 118 111 235 II 393 II

II 73 II n C§IJ 64 69 52 c .~ ~ Gil y 59 y 17i5 1]· BH i. a X y y i:i k h ]IP]U HP 5 37 116 116 99 187 49 96 76 76 254 292 158 215 187 221 I 622 I

57 s? C§:§] 60 l17ell 43 36 kh d. a X i x I x i:i 11plu JM j. a X X X i: ]DP AZ 91 79 75 109 109 489 83 36 67 91 91 260 342 91 263 II 598 I 83 194 I 693 I

52 53 56 ~8 51 l17sll6s 64 65 [§iJ 70 e X i x x i k' ] PJ k. a X X I X i: k h ]DP BK . 74 79 97 97 56 38 w 75 120 58 96 96 224 343 2 50 191 75 274 II 663 II

56 sa[§] 60 II ~411s6 f. a X i X I X 1: I kh 1u AE 64 70 66 114 114 128 224 64 2se II 466 II

161 49 43 44 ~48 45 46 49 52 44 34

9a. X i X X i 1 i t a n n a: DP AZ 44 116 64 64 63 66 107 140 39 91 224 127 173 270 0.355

69 168 70 69 75 71 74 72 1[!2] 74 74 67 1 c. a X i X I X i 1 i t a n I n a: kh ] DP EF 1[2])] 86 73 67 67 81 39 60 120 86 108 108 [12sl 458 1 195 226 148 99 314 II 691 II 0.375

69 63 66 66 I[!J]67 [ZIJ 62 59 52 54 d. a I x i x x i 1 i k' J., Sl g. a X J[ 111461162 59 104 104 62 86 [K] 72 73 73 r-----:--:-- 146 1 225 166 234 0.395 0.455

73 171 73 66 I(Z]]71 [ill 62 1 62 59 [ill 61 1[2] 68 67 70 1 e. a X i X X i 1 i k' J., BK h. a X i x x i 1 i m ] JM ll76ll100[fl] 95 95 53 61 69 59 1 85 92 74 103 103[[!1]104 31 244 1"' 76 1 268 148 189 85 269 200 I 379 0.415 0.475

69 163 66 66 [Z!] 67 [Z!J 62 JM i. a X i x x i 1 i k' J., SI ll118lj75 l107 l135 135 88 76 66 99 118 1 317 223 241 0. 545

162 69 68 68 64 IC[]68 68 62 1dB 66 68 65 II nl74 71 61 a X f V l: kh l 1Oa. X i y y ~ v i k' I· HP g. a X x oP EF 115 23 76 71 71 102 50 61 65 rs 135 41 64 99 99 140 66 247 313 115 170 173 176 135 204 239 II 626 I 0.345 0.45 5 I

71 68 70 66 II 75 IJ 64 68 65 61 60 55 1 Ga [64] 59 I 66 161 b. a X X X v i k' w LI h. a X i X X f v l: I kh ]DP PJ 176 32 85 69 69 104 60 63 77 73 102 103 95 95 67 97 233 196 176 186 173 200 73 300 162 I 526 I 0.365 0.4 65

66 60 67 61 [IT}J 66JI6s.4! 60 60 61 [ill 58 I 72 rG 65 54 c. a X i y y v i k' l w HW I. a X X X f v I: kh ]DP]U JM 94 90 100 63 63 77 63 91 56 118 86 96 92 92 109 106 223 536 94 253 140 210 118 274 201 I 865 I 0.395 0.475 I

71 63 63 65 I 73 rG 67 54 j. JI75 . 8IJ 59 69 56 ~ 67 70 52 d. a X X X f v I kh ] D~ U Ll a X i X X f V I: kh ]DP]U BH 290 18 87 82 82 115 85 155 149 150 45 122 87 87 118 71 174 230 290 187 197 II 389 I 150 254 205 475 0.395 I 0.485

57 59 56 62 61 72 65 [TI] 64 !CZp! 66 69 66 X X X v l: kh ]DP JI k. a X i X X v I: kh l 100 51 95 95 78 88 311 269 87 86 57 122 122 OITI 55 260 311 246 173 I 668 I 87 265 2 23 I 626 I 0.425 0.495

[1l]71 68 64 68 46 71 65 []J 78 ~ 74 78 64 f. a X i vlv v l:f y' ]IP]U HP I. a X i X X f v I: kh ]DP]U BK 157 47 91 90 90 116 50 171 143 83 90 157 79 79 93 59 291 253 157 228 206 I 364 I 83 326 172 I 603 I 0.445 0.4 95

163 75 75([] 69 67 71 72 64 69 55 dB [[) 66 71 65 74 69 73 66 72 !Ia. a X i X X i n iu t i:i kh I~ U BH g. a X i X X n iu y u l. BH 72 44 78 62 62 39 97 194 27 368 141 ms 108 82 76 68 68 84 42 12 2 14 84 72 184 181 291 528 188 226 152 164 98 8.295 8.385

[H 63 62 56 58 54 56 53 53 54 (]J 68 [}[ 54 56 47 b. a X i x x i n iu t i: kh lorl u KT h. a X i x x i n iu t l: kh lu PJ 155 68 68 62 62 55 93 131 77 187 278 96 187 42 72 72 181 31 198 58 195 163 155 198 117 224 534 96 221 173 221 488 8.315 8.395

1!!]]69 70 62 71 73 [[]1 66 69 58 61.9 56(§ 58 61 68 61.8 57 58 56 c. a y i x x i n iu t i:i kh l o~ u EF i. a X i x x n iu t i k J. JI 126 46 98 69 69 78 96 199 35 132 297 98 83 84 73 73 87 47 176 54 188 48 126 205 139 295 464 98 248 168 223 282 0.345 8.485

59 56 65 62 78 78(]J65 67 63 71 67 68 72 72 69 75 69 71 73[ill . a X X n iu t d. a x i x x i n iu I t i:i kh o~ u HW J. X i i dD a:i lcl1JuB H 81 67 78 67 67 74 91 152 69 175 156 101 57 67 189 109 83 54 171 58 59 78 516 81 212 141 243 1 40e I 181 233 137 225 189 DEJ 8.355 8.43

66 69.8 65 69 69([]64 68 67 69 .8 66 67 78 56 68 67 71 .7 71 72.2 68 68 64 x i x x i n iu t i k u I u l. SI k. a x i x x i n it1 t i:: kh liPlu SI 87 78 81 77 77 38 77 188 45 45 95 55 52 92 143 95 95 159 159 89 126 483 118 463 539 87 236 115 25 7 98 158 144 143 349 248 529 1112 0.355 8.685

65 68 64 68 68 68~9 63 61 65 f. a X i x x i n iu t i: kh ] 0~ u KT 115 47 86 77 77 65 89 188 65 135 412 115 210 142 m 1 612 I 0.355

164 78 66 67 69 [CZIJ] 7 8 71 . 8 49 dB 12a. a X i y i t ui n a k' Jw Ll 156 188 87 7 8 128 51 189 1 45 96 64 ms 156 187 "198~ 168 385 8.335 64 61 65 64 66 78 II 7 1 65 C§:D 66 71.7 68~6~ b . a X i i n i u t i c [ Y u: n i: I t J,t Sl X X II 115 54 83 61 I 61 6 7 45 1 42 88 48 93 93 146 82 126 188 115 198 128 187 213 239 316 8.335

lli[]] 62 65 56 67.8 68 67 I . 5 59J.I8 c . a i I 1 c[ I u J.,:t BH X X i 82 72 64 I 64 75 68 8 7 82 65 82 I ~ 223 139 155 147 8.365

d . 68 67 67 63 I 78.6 68 1! 71 67 63 63 65 56 63 66 67 65 a i I i u u lJ i: t u X y y i I Y II i y' ]w 139 58 81 7 9 I 79 74 74 1 46 86 86 42 57 118 53 57 83 83 88 98 139 218 153 3 86 128 175 193 269 8.365 I

e. 66 63 69 6 65 ~69 c:::TI"::J 68 1 a X i X X i V i:I ll J,P) U Vl 87 7 4 65 81 1 81 89 128 5 12 153 8 7 228 I 1 7 8 II 7 85 li 8.395 I 7 8 64 67 66 7 1 7 1 69 5 7 1 f. a X i X X i v i a: k" lo.Ju SI 75 78 94 81 1 81 8 7 39 1 74 141 387 7 5 245 " ~1 68 3 86 538 II 8.415 68 57 69 64 I 68 71 1 7 g . a X i i y l 6n' ~a t sr 141 78 99 168X I 84X 93 86 8 33 75 141 3 37 1 77 1 7 3 188 8.435 h . 7 8 65 c:::::z:o 6 7 7 8 . 8 68 11 7 3 68 78.6 a: X i i X i v a t Sl 147 92 98 77X I 7X7 184 55 1 16 35 95 147 267 181 171 138 8 . 455

i . 71. 5 68 71.7 68 7 1.4 69 7 2 a X i I i Sl X X i 188 84 116 188 I188 9 7 8 7 6 188 388 197 151 8 . 585

7 1 66 1 72.4 1 62 I 72 . 5 72 il 7 3 II 71 j . a: X i: lJ ( n Sl X X i I 162 168 144 1 2 4 I124 1 2 1 145 7 8 116 162 4 2 8 245 3 39 8.675

69 7 8 17 1. 771 61 71 69 7 1. 7 1 7t2 k . a X i X X i X u ~ 5: IJ Sl 153 137 147 161 161 116 137 1 95 187 116 143 153 445 277 332 366 8.725

165 APPENDIX F: The sequence /na:na/

65 70[ZI]J 71 72 .2 6811 74 II dB 63 (13)a. a n a: n a y a ]OJ TK f. a n a: n a y a: ]DP PA 34 67 98 49 81 38 58 ms 35 77 138 59 68 27 259 0. 305 0.345 0. 4351 0. 665 J

75 7511 77 1175 69 72 73 75 b. a n a: n a y 1 t BH g. a n a: n a y a: ]stutter BH 83 57 107 47 86 30 49 98 51 120 75 97 57 22 4 0.305 0.345 0.4651 0. 7451

68 69 .7 67 69 nil 71 IIGs 69 66 61 c. a n a: n a y a ]OJ BH h. a n a: n a y a: ]DP MK 81 71 93 54 78 48 78 90 63 154 82 72 61 124 0.305 0. 375 0. 5651 0. 6551

70 72 77 n [ill75ll 81 II 69. 5 nil 71 lin 69 . s 68 69 .2 d. a n a: n a y a t JD I. a n a: n a y a t MN 70 56 141 53 63 34 84 95 84 163 63 69 41 109 0.315 0.385 0. 505 J 0. 6251

64 641166. all e. a n a: n a y a ]OJ DF 77 81 134 61 63 38 58 0.345 0. 5151

166